diff --git a/docs/submission/main.aux b/docs/submission/main.aux new file mode 100644 index 0000000..229c710 --- /dev/null +++ b/docs/submission/main.aux @@ -0,0 +1,286 @@ +\relax +\providecommand\hyper@newdestlabel[2]{} +\providecommand*\HyPL@Entry[1]{} +\abx@aux@refcontext{none/global//global/global} +\abx@aux@cite{0}{scaramuzzaVisualOdometryTutorial2011} +\abx@aux@segm{0}{0}{scaramuzzaVisualOdometryTutorial2011} +\abx@aux@cite{0}{durrant-whyteSimultaneousLocalizationMapping2006} +\abx@aux@segm{0}{0}{durrant-whyteSimultaneousLocalizationMapping2006} +\abx@aux@cite{0}{davisonMonoSLAMRealtimeSingle2007} +\abx@aux@segm{0}{0}{davisonMonoSLAMRealtimeSingle2007} +\abx@aux@cite{0}{kerlDenseVisualSLAM2013} +\abx@aux@segm{0}{0}{kerlDenseVisualSLAM2013} +\abx@aux@cite{0}{newcombeKinectfusionRealtimeDense2011} +\abx@aux@segm{0}{0}{newcombeKinectfusionRealtimeDense2011} +\abx@aux@cite{0}{rusinkiewiczEfficientVariantsICP2001} +\abx@aux@segm{0}{0}{rusinkiewiczEfficientVariantsICP2001} +\abx@aux@cite{0}{mildenhallNeRFRepresentingScenes2022} +\abx@aux@segm{0}{0}{mildenhallNeRFRepresentingScenes2022} +\abx@aux@cite{0}{sandstromPointslamDenseNeural2023} +\abx@aux@segm{0}{0}{sandstromPointslamDenseNeural2023} +\abx@aux@cite{0}{sucarImapImplicitMapping2021} +\abx@aux@segm{0}{0}{sucarImapImplicitMapping2021} +\abx@aux@cite{0}{zhuNiceslamNeuralImplicit2022} +\abx@aux@segm{0}{0}{zhuNiceslamNeuralImplicit2022} +\abx@aux@cite{0}{garbinFastnerfHighfidelityNeural2021} +\abx@aux@segm{0}{0}{garbinFastnerfHighfidelityNeural2021} +\abx@aux@cite{0}{kerbl3DGaussianSplatting2023} +\abx@aux@segm{0}{0}{kerbl3DGaussianSplatting2023} +\abx@aux@cite{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@segm{0}{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@cite{0}{haRGBDGSICPSLAM2024} +\abx@aux@segm{0}{0}{haRGBDGSICPSLAM2024} +\abx@aux@cite{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@segm{0}{0}{yugayGaussianSLAMPhotorealisticDense2024} +\HyPL@Entry{0<>} +\providecommand \oddpage@label [2]{} +\@writefile{toc}{\contentsline {section}{\numberline {1}Introduction}{1}{section.1}\protected@file@percent } +\newlabel{introduction}{{1}{1}{Introduction}{section.1}{}} +\abx@aux@cite{0}{mur-artalOrbslam2OpensourceSlam2017} +\abx@aux@segm{0}{0}{mur-artalOrbslam2OpensourceSlam2017} +\abx@aux@cite{0}{camposOrbslam3AccurateOpensource2021} +\abx@aux@segm{0}{0}{camposOrbslam3AccurateOpensource2021} +\abx@aux@cite{0}{gauglitzEvaluationInterestPoint2011} +\abx@aux@segm{0}{0}{gauglitzEvaluationInterestPoint2011} +\abx@aux@cite{0}{engelDirectSparseOdometry2017} +\abx@aux@segm{0}{0}{engelDirectSparseOdometry2017} +\abx@aux@cite{0}{kerlDenseVisualSLAM2013} +\abx@aux@segm{0}{0}{kerlDenseVisualSLAM2013} +\abx@aux@cite{0}{kerlRobustOdometryEstimation2013} +\abx@aux@segm{0}{0}{kerlRobustOdometryEstimation2013} +\abx@aux@cite{0}{newcombeDTAMDenseTracking2011} +\abx@aux@segm{0}{0}{newcombeDTAMDenseTracking2011} +\abx@aux@cite{0}{whelanElasticFusionRealtimeDense2016} +\abx@aux@segm{0}{0}{whelanElasticFusionRealtimeDense2016} +\abx@aux@cite{0}{kerlRobustOdometryEstimation2013} +\abx@aux@segm{0}{0}{kerlRobustOdometryEstimation2013} +\abx@aux@cite{0}{mildenhallNeRFRepresentingScenes2022} +\abx@aux@segm{0}{0}{mildenhallNeRFRepresentingScenes2022} +\abx@aux@cite{0}{yen-chenInerfInvertingNeural2021} +\abx@aux@segm{0}{0}{yen-chenInerfInvertingNeural2021} +\abx@aux@cite{0}{mullerInstantNeuralGraphics2022} +\abx@aux@segm{0}{0}{mullerInstantNeuralGraphics2022} +\abx@aux@cite{0}{yuPlenoctreesRealtimeRendering2021} +\abx@aux@segm{0}{0}{yuPlenoctreesRealtimeRendering2021} +\abx@aux@cite{0}{fridovich-keilPlenoxelsRadianceFields2022} +\abx@aux@segm{0}{0}{fridovich-keilPlenoxelsRadianceFields2022} +\@writefile{toc}{\contentsline {section}{\numberline {2}Related Work}{2}{section.2}\protected@file@percent } +\newlabel{related-work}{{2}{2}{Related Work}{section.2}{}} +\@writefile{toc}{\contentsline {subsection}{\numberline {2.1}Classical RGB-D Localization}{2}{subsection.2.1}\protected@file@percent } +\newlabel{classical-rgb-d-localization}{{2.1}{2}{Classical RGB-D Localization}{subsection.2.1}{}} +\@writefile{toc}{\contentsline {subsection}{\numberline {2.2}NeRF-Based Localization}{2}{subsection.2.2}\protected@file@percent } +\newlabel{nerf-based-localization}{{2.2}{2}{NeRF-Based Localization}{subsection.2.2}{}} +\abx@aux@cite{0}{kerbl3DGaussianSplatting2023} +\abx@aux@segm{0}{0}{kerbl3DGaussianSplatting2023} +\abx@aux@cite{0}{keethaSplaTAMSplatTrack2024} +\abx@aux@segm{0}{0}{keethaSplaTAMSplatTrack2024} +\abx@aux@cite{0}{huCGSLAMEfficientDense2024} +\abx@aux@segm{0}{0}{huCGSLAMEfficientDense2024} +\abx@aux@cite{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@segm{0}{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@cite{0}{haRGBDGSICPSLAM2024} +\abx@aux@segm{0}{0}{haRGBDGSICPSLAM2024} +\abx@aux@cite{0}{beslMethodRegistration3shapes1992} +\abx@aux@segm{0}{0}{beslMethodRegistration3shapes1992} +\abx@aux@cite{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@segm{0}{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@cite{0}{segalGeneralizedicp2009a} +\abx@aux@segm{0}{0}{segalGeneralizedicp2009a} +\abx@aux@cite{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@segm{0}{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@cite{0}{parkColoredPointCloud2017} +\abx@aux@segm{0}{0}{parkColoredPointCloud2017} +\abx@aux@cite{0}{steinbruckerRealtimeVisualOdometry2011} +\abx@aux@segm{0}{0}{steinbruckerRealtimeVisualOdometry2011} +\abx@aux@cite{0}{pomerleauComparingICPVariants2013} +\abx@aux@segm{0}{0}{pomerleauComparingICPVariants2013} +\abx@aux@cite{0}{kuipersQuaternionsRotationSequences1999} +\abx@aux@segm{0}{0}{kuipersQuaternionsRotationSequences1999} +\abx@aux@cite{0}{mildenhallNeRFRepresentingScenes2022} +\abx@aux@segm{0}{0}{mildenhallNeRFRepresentingScenes2022} +\abx@aux@cite{0}{kerbl3DGaussianSplatting2023} +\abx@aux@segm{0}{0}{kerbl3DGaussianSplatting2023} +\@writefile{toc}{\contentsline {subsection}{\numberline {2.3}Gaussian-Based Localization}{3}{subsection.2.3}\protected@file@percent } +\newlabel{gaussian-based-localization}{{2.3}{3}{Gaussian-Based Localization}{subsection.2.3}{}} +\@writefile{toc}{\contentsline {section}{\numberline {3}Method}{3}{section.3}\protected@file@percent } +\newlabel{method}{{3}{3}{Method}{section.3}{}} +\abx@aux@cite{0}{kerbl3DGaussianSplatting2023} +\abx@aux@segm{0}{0}{kerbl3DGaussianSplatting2023} +\abx@aux@cite{0}{kerbl3DGaussianSplatting2023} +\abx@aux@segm{0}{0}{kerbl3DGaussianSplatting2023} +\abx@aux@cite{0}{zwickerEWASplatting2002} +\abx@aux@segm{0}{0}{zwickerEWASplatting2002} +\abx@aux@cite{0}{kerbl3DGaussianSplatting2023} +\abx@aux@segm{0}{0}{kerbl3DGaussianSplatting2023} +\@writefile{lof}{\contentsline {figure}{\numberline {1}{\ignorespaces We propose \textbf {GSplatLoc}, a novel camera localization method that leverages the differentiable rendering capabilities of 3D Gaussian splatting for efficient and accurate pose estimation.}}{4}{figure.caption.1}\protected@file@percent } +\providecommand*\caption@xref[2]{\@setref\relax\@undefined{#1}} +\newlabel{fig:cross-column-image}{{1}{4}{We propose \textbf {GSplatLoc}, a novel camera localization method that leverages the differentiable rendering capabilities of 3D Gaussian splatting for efficient and accurate pose estimation}{figure.caption.1}{}} +\@writefile{toc}{\contentsline {subsection}{\numberline {3.1}Scene Representation}{4}{subsection.3.1}\protected@file@percent } +\newlabel{scene-representation}{{3.1}{4}{Scene Representation}{subsection.3.1}{}} +\@writefile{toc}{\contentsline {subsection}{\numberline {3.2}Depth Rendering}{4}{subsection.3.2}\protected@file@percent } +\newlabel{depth-rendering}{{3.2}{4}{Depth Rendering}{subsection.3.2}{}} +\abx@aux@cite{0}{kuipersQuaternionsRotationSequences1999} +\abx@aux@segm{0}{0}{kuipersQuaternionsRotationSequences1999} +\abx@aux@cite{0}{kanopoulosDesignImageEdge1988} +\abx@aux@segm{0}{0}{kanopoulosDesignImageEdge1988} +\@writefile{toc}{\contentsline {subsection}{\numberline {3.3}Localization as Image Alignment}{5}{subsection.3.3}\protected@file@percent } +\newlabel{localization-as-image-alignment}{{3.3}{5}{Localization as Image Alignment}{subsection.3.3}{}} +\abx@aux@cite{0}{kingmaAdamMethodStochastic2014} +\abx@aux@segm{0}{0}{kingmaAdamMethodStochastic2014} +\@writefile{toc}{\contentsline {subsection}{\numberline {3.4}Pipeline}{6}{subsection.3.4}\protected@file@percent } +\newlabel{pipeline}{{3.4}{6}{Pipeline}{subsection.3.4}{}} +\@writefile{toc}{\contentsline {section}{\numberline {4}Evaluation}{6}{section.4}\protected@file@percent } +\newlabel{evaluation}{{4}{6}{Evaluation}{section.4}{}} +\@writefile{toc}{\contentsline {subsection}{\numberline {4.1}Experimental Setup}{6}{subsection.4.1}\protected@file@percent } +\newlabel{experimental-setup}{{4.1}{6}{Experimental Setup}{subsection.4.1}{}} +\abx@aux@cite{0}{straubReplicaDatasetDigital2019} +\abx@aux@segm{0}{0}{straubReplicaDatasetDigital2019} +\abx@aux@cite{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@segm{0}{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@cite{0}{sucarImapImplicitMapping2021} +\abx@aux@segm{0}{0}{sucarImapImplicitMapping2021} +\abx@aux@cite{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@segm{0}{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@cite{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@segm{0}{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@cite{0}{haRGBDGSICPSLAM2024} +\abx@aux@segm{0}{0}{haRGBDGSICPSLAM2024} +\abx@aux@cite{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@segm{0}{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@cite{0}{straubReplicaDatasetDigital2019} +\abx@aux@segm{0}{0}{straubReplicaDatasetDigital2019} +\abx@aux@cite{0}{straubReplicaDatasetDigital2019} +\abx@aux@segm{0}{0}{straubReplicaDatasetDigital2019} +\abx@aux@cite{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@segm{0}{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@cite{0}{haRGBDGSICPSLAM2024} +\abx@aux@segm{0}{0}{haRGBDGSICPSLAM2024} +\abx@aux@cite{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@segm{0}{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@cite{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@segm{0}{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@cite{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@segm{0}{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@cite{0}{straubReplicaDatasetDigital2019} +\abx@aux@segm{0}{0}{straubReplicaDatasetDigital2019} +\abx@aux@cite{0}{straubReplicaDatasetDigital2019} +\abx@aux@segm{0}{0}{straubReplicaDatasetDigital2019} +\abx@aux@cite{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@segm{0}{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@cite{0}{haRGBDGSICPSLAM2024} +\abx@aux@segm{0}{0}{haRGBDGSICPSLAM2024} +\abx@aux@cite{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@segm{0}{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@cite{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@segm{0}{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@cite{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@segm{0}{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@cite{0}{haRGBDGSICPSLAM2024} +\abx@aux@segm{0}{0}{haRGBDGSICPSLAM2024} +\abx@aux@cite{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@segm{0}{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@cite{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@segm{0}{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@cite{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@segm{0}{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@cite{0}{haRGBDGSICPSLAM2024} +\abx@aux@segm{0}{0}{haRGBDGSICPSLAM2024} +\abx@aux@cite{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@segm{0}{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@cite{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@segm{0}{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@cite{0}{haRGBDGSICPSLAM2024} +\abx@aux@segm{0}{0}{haRGBDGSICPSLAM2024} +\abx@aux@cite{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@segm{0}{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@cite{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@segm{0}{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@cite{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@segm{0}{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@cite{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@segm{0}{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@cite{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@segm{0}{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@cite{0}{haRGBDGSICPSLAM2024} +\abx@aux@segm{0}{0}{haRGBDGSICPSLAM2024} +\abx@aux@cite{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@segm{0}{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@cite{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@segm{0}{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@cite{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@segm{0}{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@cite{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@segm{0}{0}{pengRTGSLAMRealtime3D2024} +\abx@aux@cite{0}{haRGBDGSICPSLAM2024} +\abx@aux@segm{0}{0}{haRGBDGSICPSLAM2024} +\abx@aux@cite{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@segm{0}{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@cite{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@segm{0}{0}{yugayGaussianSLAMPhotorealisticDense2024} +\abx@aux@cite{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@segm{0}{0}{sturmBenchmarkEvaluationRGBD2012} +\@writefile{toc}{\contentsline {subsection}{\numberline {4.2}Localization Evaluation}{7}{subsection.4.2}\protected@file@percent } +\newlabel{localization-evaluation}{{4.2}{7}{Localization Evaluation}{subsection.4.2}{}} +\@writefile{lot}{\contentsline {table}{\numberline {\textbf {1}}{\ignorespaces \textbf {Replica\blx@tocontentsinit {0}\cite {straubReplicaDatasetDigital2019} (ATE RMSE ↓[cm]).}}}{7}{table.caption.2}\protected@file@percent } +\newlabel{table:_textbf_replica_cite}{{\textbf {1}}{7}{\textbf {Replica\cite {straubReplicaDatasetDigital2019} (ATE RMSE ↓[cm]).}}{table.caption.2}{}} +\@writefile{lot}{\contentsline {table}{\numberline {\textbf {2}}{\ignorespaces \textbf {Replica\blx@tocontentsinit {0}\cite {straubReplicaDatasetDigital2019} (AAE RMSE ↓[°]).}}}{7}{table.caption.3}\protected@file@percent } +\newlabel{table:_textbf_replica_cite}{{\textbf {2}}{7}{\textbf {Replica\cite {straubReplicaDatasetDigital2019} (AAE RMSE ↓[°]).}}{table.caption.3}{}} +\@writefile{lot}{\contentsline {table}{\numberline {\textbf {3}}{\ignorespaces \textbf {TUM\blx@tocontentsinit {0}\cite {sturmBenchmarkEvaluationRGBD2012} (ATE RMSE ↓[cm]).}}}{7}{table.caption.4}\protected@file@percent } +\newlabel{table:_textbf_tum_cite_stu}{{\textbf {3}}{7}{\textbf {TUM\cite {sturmBenchmarkEvaluationRGBD2012} (ATE RMSE ↓[cm]).}}{table.caption.4}{}} +\abx@aux@cite{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@segm{0}{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@cite{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@segm{0}{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@cite{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@segm{0}{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@cite{0}{sturmBenchmarkEvaluationRGBD2012} +\abx@aux@segm{0}{0}{sturmBenchmarkEvaluationRGBD2012} +\@writefile{lot}{\contentsline {table}{\numberline {\textbf {4}}{\ignorespaces \textbf {TUM\blx@tocontentsinit {0}\cite {sturmBenchmarkEvaluationRGBD2012} (AAE RMSE ↓[°]).}}}{8}{table.caption.5}\protected@file@percent } +\newlabel{table:_textbf_tum_cite_stu}{{\textbf {4}}{8}{\textbf {TUM\cite {sturmBenchmarkEvaluationRGBD2012} (AAE RMSE ↓[°]).}}{table.caption.5}{}} +\@writefile{toc}{\contentsline {subsection}{\numberline {4.3}Discussion}{8}{subsection.4.3}\protected@file@percent } +\newlabel{discussion}{{4.3}{8}{Discussion}{subsection.4.3}{}} +\@writefile{toc}{\contentsline {subsection}{\numberline {4.4}Limitations}{8}{subsection.4.4}\protected@file@percent } +\newlabel{limitations}{{4.4}{8}{Limitations}{subsection.4.4}{}} +\@writefile{toc}{\contentsline {section}{\numberline {5}Conclusion}{8}{section.5}\protected@file@percent } +\newlabel{conclusion}{{5}{8}{Conclusion}{section.5}{}} +\abx@aux@read@bbl@mdfivesum{1360600842772F54FA94B5276D91E24F} +\abx@aux@defaultrefcontext{0}{scaramuzzaVisualOdometryTutorial2011}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{durrant-whyteSimultaneousLocalizationMapping2006}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{davisonMonoSLAMRealtimeSingle2007}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{kerlDenseVisualSLAM2013}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{newcombeKinectfusionRealtimeDense2011}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{rusinkiewiczEfficientVariantsICP2001}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{mildenhallNeRFRepresentingScenes2022}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{sandstromPointslamDenseNeural2023}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{sucarImapImplicitMapping2021}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{zhuNiceslamNeuralImplicit2022}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{garbinFastnerfHighfidelityNeural2021}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{kerbl3DGaussianSplatting2023}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{pengRTGSLAMRealtime3D2024}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{haRGBDGSICPSLAM2024}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{yugayGaussianSLAMPhotorealisticDense2024}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{mur-artalOrbslam2OpensourceSlam2017}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{camposOrbslam3AccurateOpensource2021}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{gauglitzEvaluationInterestPoint2011}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{engelDirectSparseOdometry2017}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{kerlRobustOdometryEstimation2013}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{newcombeDTAMDenseTracking2011}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{whelanElasticFusionRealtimeDense2016}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{yen-chenInerfInvertingNeural2021}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{mullerInstantNeuralGraphics2022}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{yuPlenoctreesRealtimeRendering2021}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{fridovich-keilPlenoxelsRadianceFields2022}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{keethaSplaTAMSplatTrack2024}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{huCGSLAMEfficientDense2024}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{beslMethodRegistration3shapes1992}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{segalGeneralizedicp2009a}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{parkColoredPointCloud2017}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{steinbruckerRealtimeVisualOdometry2011}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{pomerleauComparingICPVariants2013}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{kuipersQuaternionsRotationSequences1999}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{zwickerEWASplatting2002}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{kanopoulosDesignImageEdge1988}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{kingmaAdamMethodStochastic2014}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{straubReplicaDatasetDigital2019}{none/global//global/global} +\abx@aux@defaultrefcontext{0}{sturmBenchmarkEvaluationRGBD2012}{none/global//global/global} +\gdef \@abspage@last{11} diff --git a/docs/submission/main.bbl b/docs/submission/main.bbl new file mode 100644 index 0000000..223ce88 --- /dev/null +++ b/docs/submission/main.bbl @@ -0,0 +1,2496 @@ +% $ biblatex auxiliary file $ +% $ biblatex bbl format version 3.2 $ +% Do not modify the above lines! +% +% This is an auxiliary file used by the 'biblatex' package. +% This file may safely be deleted. It will be recreated by +% biber as required. +% +\begingroup +\makeatletter +\@ifundefined{ver@biblatex.sty} + {\@latex@error + {Missing 'biblatex' package} + {The bibliography requires the 'biblatex' package.} + \aftergroup\endinput} + {} +\endgroup + + +\refsection{0} + \datalist[entry]{none/global//global/global} + \entry{scaramuzzaVisualOdometryTutorial2011}{article}{} + \name{author}{2}{}{% + {{hash=dac21ab4ede215439fcc6b051be53a11}{% + family={Scaramuzza}, + familyi={S\bibinitperiod}, + given={Davide}, + giveni={D\bibinitperiod}}}% + {{hash=ea987af098c553e56730615d244f1dd6}{% + family={Fraundorfer}, + familyi={F\bibinitperiod}, + given={Friedrich}, + giveni={F\bibinitperiod}}}% + } + \list{publisher}{1}{% + {IEEE}% + } + \strng{namehash}{e3fadefef92a28fcd8544db3cc769ba2} + \strng{fullhash}{e3fadefef92a28fcd8544db3cc769ba2} + \strng{bibnamehash}{e3fadefef92a28fcd8544db3cc769ba2} + \strng{authorbibnamehash}{e3fadefef92a28fcd8544db3cc769ba2} + \strng{authornamehash}{e3fadefef92a28fcd8544db3cc769ba2} + \strng{authorfullhash}{e3fadefef92a28fcd8544db3cc769ba2} + \field{sortinit}{1} + \field{sortinithash}{4f6aaa89bab872aa0999fec09ff8e98a} + \field{labelnamesource}{author} + \field{labeltitlesource}{title} + \field{journaltitle}{IEEE robotics \& automation magazine} + \field{number}{4} + \field{title}{Visual Odometry [Tutorial]} + \field{urlday}{26} + \field{urlmonth}{9} + \field{urlyear}{2024} + \field{volume}{18} + \field{year}{2011} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{80\bibrangedash 92} + \range{pages}{13} + \verb{doi} + \verb 10.1109/MRA.2011.943233 + \endverb + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\introduce\Scaramuzza_Fraundorfer_2011_Visual odometry [tutorial].pdf + \endverb + \verb{urlraw} + \verb https://ieeexplore.ieee.org/abstract/document/6096039/ + \endverb + \verb{url} + \verb https://ieeexplore.ieee.org/abstract/document/6096039/ + \endverb + \endentry + \entry{durrant-whyteSimultaneousLocalizationMapping2006}{article}{} + \name{author}{2}{}{% + {{hash=25adb16e1e973b04b027303ccb7e5fcf}{% + family={Durrant-Whyte}, + familyi={D\bibinithyphendelim W\bibinitperiod}, + given={Hugh}, + giveni={H\bibinitperiod}}}% + {{hash=a37c63e1fc8eaedec4ea9e27569c27d3}{% + family={Bailey}, + familyi={B\bibinitperiod}, + given={Tim}, + giveni={T\bibinitperiod}}}% + } + \list{publisher}{1}{% + {IEEE}% + } + \strng{namehash}{210de09a158ffbef02c16824a43de295} + \strng{fullhash}{210de09a158ffbef02c16824a43de295} + \strng{bibnamehash}{210de09a158ffbef02c16824a43de295} + \strng{authorbibnamehash}{210de09a158ffbef02c16824a43de295} + \strng{authornamehash}{210de09a158ffbef02c16824a43de295} + \strng{authorfullhash}{210de09a158ffbef02c16824a43de295} + \field{sortinit}{2} + \field{sortinithash}{8b555b3791beccb63322c22f3320aa9a} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{journaltitle}{IEEE robotics \& automation magazine} + \field{number}{2} + \field{shorttitle}{Simultaneous Localization and Mapping} + \field{title}{Simultaneous Localization and Mapping: Part {{I}}} + \field{urlday}{26} + \field{urlmonth}{9} + \field{urlyear}{2024} + \field{volume}{13} + \field{year}{2006} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{99\bibrangedash 110} + \range{pages}{12} + \verb{doi} + \verb 10.1109/MRA.2006.1638022 + \endverb + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\introduce\Durrant-Whyte_Bailey_2006_Simultaneous localization and mapping.pdf + \endverb + \verb{urlraw} + \verb https://ieeexplore.ieee.org/abstract/document/1638022/ + \endverb + \verb{url} + \verb https://ieeexplore.ieee.org/abstract/document/1638022/ + \endverb + \endentry + \entry{davisonMonoSLAMRealtimeSingle2007}{article}{} + \name{author}{4}{}{% + {{hash=6800da0b1eba1591ea0c413f8efceae3}{% + family={Davison}, + familyi={D\bibinitperiod}, + given={Andrew\bibnamedelima J.}, + giveni={A\bibinitperiod\bibinitdelim J\bibinitperiod}}}% + {{hash=ee8d212ba96ab9b80423488534d7574c}{% + family={Reid}, + familyi={R\bibinitperiod}, + given={Ian\bibnamedelima D.}, + giveni={I\bibinitperiod\bibinitdelim D\bibinitperiod}}}% + {{hash=45c6c69807f417e619d3d397c0448d50}{% + family={Molton}, + familyi={M\bibinitperiod}, + given={Nicholas\bibnamedelima D.}, + giveni={N\bibinitperiod\bibinitdelim D\bibinitperiod}}}% + {{hash=d976a252b6fa0efe1839e7c6de032d43}{% + family={Stasse}, + familyi={S\bibinitperiod}, + given={Olivier}, + giveni={O\bibinitperiod}}}% + } + \list{publisher}{1}{% + {IEEE}% + } + \strng{namehash}{fceadd62e12f3daea401c47719d984dc} + \strng{fullhash}{d34af9fb5c054d89ccbb69195f280f82} + \strng{bibnamehash}{d34af9fb5c054d89ccbb69195f280f82} + \strng{authorbibnamehash}{d34af9fb5c054d89ccbb69195f280f82} + \strng{authornamehash}{fceadd62e12f3daea401c47719d984dc} + \strng{authorfullhash}{d34af9fb5c054d89ccbb69195f280f82} + \field{sortinit}{2} + \field{sortinithash}{8b555b3791beccb63322c22f3320aa9a} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{journaltitle}{IEEE transactions on pattern analysis and machine intelligence} + \field{number}{6} + \field{shorttitle}{{{MonoSLAM}}} + \field{title}{{{MonoSLAM}}: {{Real-time}} Single Camera {{SLAM}}} + \field{urlday}{26} + \field{urlmonth}{9} + \field{urlyear}{2024} + \field{volume}{29} + \field{year}{2007} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{1052\bibrangedash 1067} + \range{pages}{16} + \verb{doi} + \verb 10.1109/TPAMI.2007.1049 + \endverb + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\introduce\Davison et al_2007_MonoSLAM.pdf + \endverb + \verb{urlraw} + \verb https://ieeexplore.ieee.org/abstract/document/4160954/ + \endverb + \verb{url} + \verb https://ieeexplore.ieee.org/abstract/document/4160954/ + \endverb + \endentry + \entry{kerlDenseVisualSLAM2013}{inproceedings}{} + \name{author}{3}{}{% + {{hash=c8b46d457d7577efc9d388d1c40fb783}{% + family={Kerl}, + familyi={K\bibinitperiod}, + given={Christian}, + giveni={C\bibinitperiod}}}% + {{hash=9e291584a3b29b27c88ecfe0a566274c}{% + family={Sturm}, + familyi={S\bibinitperiod}, + given={Jürgen}, + giveni={J\bibinitperiod}}}% + {{hash=1bd2b6b6ca2fc15a90f164070b626131}{% + family={Cremers}, + familyi={C\bibinitperiod}, + given={Daniel}, + giveni={D\bibinitperiod}}}% + } + \list{publisher}{1}{% + {IEEE}% + } + \strng{namehash}{4088c2282796db3475bd1d361a9ac3f1} + \strng{fullhash}{21c79321a177b2edd1d9ad2ca8f11e28} + \strng{bibnamehash}{21c79321a177b2edd1d9ad2ca8f11e28} + \strng{authorbibnamehash}{21c79321a177b2edd1d9ad2ca8f11e28} + \strng{authornamehash}{4088c2282796db3475bd1d361a9ac3f1} + \strng{authorfullhash}{21c79321a177b2edd1d9ad2ca8f11e28} + \field{extraname}{1} + \field{sortinit}{3} + \field{sortinithash}{ad6fe7482ffbd7b9f99c9e8b5dccd3d7} + \field{labelnamesource}{author} + \field{labeltitlesource}{title} + \field{booktitle}{2013 {{IEEE}}/{{RSJ International Conference}} on {{Intelligent Robots}} and {{Systems}}} + \field{title}{Dense Visual {{SLAM}} for {{RGB-D}} Cameras} + \field{urlday}{22} + \field{urlmonth}{8} + \field{urlyear}{2024} + \field{year}{2013} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{2100\bibrangedash 2106} + \range{pages}{7} + \verb{doi} + \verb 10.1109/IROS.2013.6696650 + \endverb + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\Kerl et al_2013_Dense visual SLAM for RGB-D cameras.pdf + \endverb + \verb{urlraw} + \verb https://ieeexplore.ieee.org/abstract/document/6696650/ + \endverb + \verb{url} + \verb https://ieeexplore.ieee.org/abstract/document/6696650/ + \endverb + \endentry + \entry{newcombeKinectfusionRealtimeDense2011}{inproceedings}{} + \name{author}{10}{}{% + {{hash=7b635dacb5ef25a6f9c500f8c62aacd8}{% + family={Newcombe}, + familyi={N\bibinitperiod}, + given={Richard\bibnamedelima A.}, + giveni={R\bibinitperiod\bibinitdelim A\bibinitperiod}}}% + {{hash=19d80a2a01693de3631decaa652d4d3f}{% + family={Izadi}, + familyi={I\bibinitperiod}, + given={Shahram}, + giveni={S\bibinitperiod}}}% + {{hash=e4282a8b050f6ec88f43896369f6fe38}{% + family={Hilliges}, + familyi={H\bibinitperiod}, + given={Otmar}, + giveni={O\bibinitperiod}}}% + {{hash=23c72b179d65cca0461f33a964f1f98b}{% + family={Molyneaux}, + familyi={M\bibinitperiod}, + given={David}, + giveni={D\bibinitperiod}}}% + {{hash=4bbde8a08938e282b6bd70ea30ba2b09}{% + family={Kim}, + familyi={K\bibinitperiod}, + given={David}, + giveni={D\bibinitperiod}}}% + {{hash=6800da0b1eba1591ea0c413f8efceae3}{% + family={Davison}, + familyi={D\bibinitperiod}, + given={Andrew\bibnamedelima J.}, + giveni={A\bibinitperiod\bibinitdelim J\bibinitperiod}}}% + {{hash=d98a66dcfa35ba76c93f9e8a73b3650c}{% + family={Kohi}, + familyi={K\bibinitperiod}, + given={Pushmeet}, + giveni={P\bibinitperiod}}}% + {{hash=3b90a1e3dad54d8cb1230eec501fdf04}{% + family={Shotton}, + familyi={S\bibinitperiod}, + given={Jamie}, + giveni={J\bibinitperiod}}}% + {{hash=16c2fad85c202f0f60f5d77854ca1cd8}{% + family={Hodges}, + familyi={H\bibinitperiod}, + given={Steve}, + giveni={S\bibinitperiod}}}% + {{hash=8eb0347090bee0e3077a241bdad85387}{% + family={Fitzgibbon}, + familyi={F\bibinitperiod}, + given={Andrew}, + giveni={A\bibinitperiod}}}% + } + \list{publisher}{1}{% + {Ieee}% + } + \strng{namehash}{1716b787f56b1649cb7cef04aa0bc5c4} + \strng{fullhash}{141c488f773c92b7e6a2650d07cf9dcb} + \strng{bibnamehash}{1716b787f56b1649cb7cef04aa0bc5c4} + \strng{authorbibnamehash}{1716b787f56b1649cb7cef04aa0bc5c4} + \strng{authornamehash}{1716b787f56b1649cb7cef04aa0bc5c4} + \strng{authorfullhash}{141c488f773c92b7e6a2650d07cf9dcb} + \field{extraname}{1} + \field{sortinit}{4} + \field{sortinithash}{9381316451d1b9788675a07e972a12a7} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{booktitle}{2011 10th {{IEEE}} International Symposium on Mixed and Augmented Reality} + \field{shorttitle}{Kinectfusion} + \field{title}{Kinectfusion: {{Real-time}} Dense Surface Mapping and Tracking} + \field{urlday}{3} + \field{urlmonth}{9} + \field{urlyear}{2024} + \field{year}{2011} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{127\bibrangedash 136} + \range{pages}{10} + \verb{doi} + \verb 10.1109/ISMAR.2011.6092378 + \endverb + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\relate_works\Newcombe et al_2011_Kinectfusion.pdf + \endverb + \verb{urlraw} + \verb https://ieeexplore.ieee.org/abstract/document/6162880/ + \endverb + \verb{url} + \verb https://ieeexplore.ieee.org/abstract/document/6162880/ + \endverb + \endentry + \entry{rusinkiewiczEfficientVariantsICP2001}{inproceedings}{} + \name{author}{2}{}{% + {{hash=7828482a1cae897b8d87b30255d79a7d}{% + family={Rusinkiewicz}, + familyi={R\bibinitperiod}, + given={Szymon}, + giveni={S\bibinitperiod}}}% + {{hash=1315bf00f5fa82534be18df443fc255c}{% + family={Levoy}, + familyi={L\bibinitperiod}, + given={Marc}, + giveni={M\bibinitperiod}}}% + } + \list{publisher}{1}{% + {IEEE}% + } + \strng{namehash}{0c03d43219d4ded30e34e6bb593488e8} + \strng{fullhash}{0c03d43219d4ded30e34e6bb593488e8} + \strng{bibnamehash}{0c03d43219d4ded30e34e6bb593488e8} + \strng{authorbibnamehash}{0c03d43219d4ded30e34e6bb593488e8} + \strng{authornamehash}{0c03d43219d4ded30e34e6bb593488e8} + \strng{authorfullhash}{0c03d43219d4ded30e34e6bb593488e8} + \field{sortinit}{4} + \field{sortinithash}{9381316451d1b9788675a07e972a12a7} + \field{labelnamesource}{author} + \field{labeltitlesource}{title} + \field{booktitle}{Proceedings Third International Conference on 3-{{D}} Digital Imaging and Modeling} + \field{title}{Efficient Variants of the {{ICP}} Algorithm} + \field{urlday}{26} + \field{urlmonth}{9} + \field{urlyear}{2024} + \field{year}{2001} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{145\bibrangedash 152} + \range{pages}{8} + \verb{doi} + \verb 10.1109/IM.2001.924423 + \endverb + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\introduce\Rusinkiewicz_Levoy_2001_Efficient variants of the ICP algorithm.pdf + \endverb + \verb{urlraw} + \verb https://ieeexplore.ieee.org/abstract/document/924423/ + \endverb + \verb{url} + \verb https://ieeexplore.ieee.org/abstract/document/924423/ + \endverb + \endentry + \entry{mildenhallNeRFRepresentingScenes2022}{article}{} + \name{author}{6}{}{% + {{hash=a3567364ab52973659adf74ee0c6e6f3}{% + family={Mildenhall}, + familyi={M\bibinitperiod}, + given={Ben}, + giveni={B\bibinitperiod}}}% + {{hash=b503d22b8c86fb9dfd477d2f70a59650}{% + family={Srinivasan}, + familyi={S\bibinitperiod}, + given={Pratul\bibnamedelima P.}, + giveni={P\bibinitperiod\bibinitdelim P\bibinitperiod}}}% + {{hash=9a18985a5c8b8be23d0ec9de0a3c87b9}{% + family={Tancik}, + familyi={T\bibinitperiod}, + given={Matthew}, + giveni={M\bibinitperiod}}}% + {{hash=562f3254a241932db4a6d44a97473ec5}{% + family={Barron}, + familyi={B\bibinitperiod}, + given={Jonathan\bibnamedelima T.}, + giveni={J\bibinitperiod\bibinitdelim T\bibinitperiod}}}% + {{hash=e7e274981ceb0dd123d44af35cc7bedf}{% + family={Ramamoorthi}, + familyi={R\bibinitperiod}, + given={Ravi}, + giveni={R\bibinitperiod}}}% + {{hash=96e30bdc2cf905aa1109c46617b16252}{% + family={Ng}, + familyi={N\bibinitperiod}, + given={Ren}, + giveni={R\bibinitperiod}}}% + } + \strng{namehash}{094342ecc1d4c7df1a909882dd0a89cd} + \strng{fullhash}{9bee638bba28fb3cd739399aa8dc9983} + \strng{bibnamehash}{9bee638bba28fb3cd739399aa8dc9983} + \strng{authorbibnamehash}{9bee638bba28fb3cd739399aa8dc9983} + \strng{authornamehash}{094342ecc1d4c7df1a909882dd0a89cd} + \strng{authorfullhash}{9bee638bba28fb3cd739399aa8dc9983} + \field{sortinit}{5} + \field{sortinithash}{20e9b4b0b173788c5dace24730f47d8c} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{abstract}{We present a method that achieves state-of-the-art results for synthesizing novel views of complex scenes by optimizing an underlying continuous volumetric scene function using a sparse set of input views. Our algorithm represents a scene using a fully connected (nonconvolutional) deep network, whose input is a single continuous 5D coordinate (spatial location ( x , y , z ) and viewing direction ( θ, ϕ )) and whose output is the volume density and view-dependent emitted radiance at that spatial location. We synthesize views by querying 5D coordinates along camera rays and use classic volume rendering techniques to project the output colors and densities into an image. Because volume rendering is naturally differentiable, the only input required to optimize our representation is a set of images with known camera poses. We describe how to effectively optimize neural radiance fields to render photorealistic novel views of scenes with complicated geometry and appearance, and demonstrate results that outperform prior work on neural rendering and view synthesis.} + \field{issn}{0001-0782, 1557-7317} + \field{journaltitle}{Communications of the ACM} + \field{langid}{english} + \field{month}{1} + \field{number}{1} + \field{shortjournal}{Commun. ACM} + \field{shorttitle}{{{NeRF}}} + \field{title}{{{NeRF}}: Representing Scenes as Neural Radiance Fields for View Synthesis} + \field{urlday}{22} + \field{urlmonth}{8} + \field{urlyear}{2024} + \field{volume}{65} + \field{year}{2022} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{99\bibrangedash 106} + \range{pages}{8} + \verb{doi} + \verb 10.1145/3503250 + \endverb + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\introduce\Mildenhall et al_2022_NeRF.pdf + \endverb + \verb{urlraw} + \verb https://dl.acm.org/doi/10.1145/3503250 + \endverb + \verb{url} + \verb https://dl.acm.org/doi/10.1145/3503250 + \endverb + \endentry + \entry{sandstromPointslamDenseNeural2023}{inproceedings}{} + \name{author}{4}{}{% + {{hash=d9eb3f671403ee2a28d9e48677737725}{% + family={Sandström}, + familyi={S\bibinitperiod}, + given={Erik}, + giveni={E\bibinitperiod}}}% + {{hash=0e1dff109644a2423d611a0bfe123178}{% + family={Li}, + familyi={L\bibinitperiod}, + given={Yue}, + giveni={Y\bibinitperiod}}}% + {{hash=d3bc7b9be5061411a87f19fcdf39499c}{% + family={Van\bibnamedelima Gool}, + familyi={V\bibinitperiod\bibinitdelim G\bibinitperiod}, + given={Luc}, + giveni={L\bibinitperiod}}}% + {{hash=44e1d63adacaa62a8b257c28eb6145e2}{% + family={Oswald}, + familyi={O\bibinitperiod}, + given={Martin\bibnamedelima R.}, + giveni={M\bibinitperiod\bibinitdelim R\bibinitperiod}}}% + } + \strng{namehash}{2bd91d8d2f82b3e665ce5d86e849ca87} + \strng{fullhash}{4e1210a52b1afbae2f89585a7117a2e5} + \strng{bibnamehash}{4e1210a52b1afbae2f89585a7117a2e5} + \strng{authorbibnamehash}{4e1210a52b1afbae2f89585a7117a2e5} + \strng{authornamehash}{2bd91d8d2f82b3e665ce5d86e849ca87} + \strng{authorfullhash}{4e1210a52b1afbae2f89585a7117a2e5} + \field{sortinit}{6} + \field{sortinithash}{b33bc299efb3c36abec520a4c896a66d} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{booktitle}{Proceedings of the {{IEEE}}/{{CVF International Conference}} on {{Computer Vision}}} + \field{shorttitle}{Point-Slam} + \field{title}{Point-Slam: {{Dense}} Neural Point Cloud-Based Slam} + \field{urlday}{22} + \field{urlmonth}{8} + \field{urlyear}{2024} + \field{year}{2023} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{18433\bibrangedash 18444} + \range{pages}{12} + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\introduce\Sandström et al_2023_Point-slam.pdf + \endverb + \verb{urlraw} + \verb http://openaccess.thecvf.com/content/ICCV2023/html/Sandstrom_Point-SLAM_Dense_Neural_Point_Cloud-based_SLAM_ICCV_2023_paper.html + \endverb + \verb{url} + \verb http://openaccess.thecvf.com/content/ICCV2023/html/Sandstrom_Point-SLAM_Dense_Neural_Point_Cloud-based_SLAM_ICCV_2023_paper.html + \endverb + \keyw{⛔ No DOI found} + \endentry + \entry{sucarImapImplicitMapping2021}{inproceedings}{} + \name{author}{4}{}{% + {{hash=8f1a5e878ee54b003a7e019a1b6b7915}{% + family={Sucar}, + familyi={S\bibinitperiod}, + given={Edgar}, + giveni={E\bibinitperiod}}}% + {{hash=6f1e887295b9ad4b9ac870667d61b091}{% + family={Liu}, + familyi={L\bibinitperiod}, + given={Shikun}, + giveni={S\bibinitperiod}}}% + {{hash=a2865884ed184115cce321b3469bc252}{% + family={Ortiz}, + familyi={O\bibinitperiod}, + given={Joseph}, + giveni={J\bibinitperiod}}}% + {{hash=6800da0b1eba1591ea0c413f8efceae3}{% + family={Davison}, + familyi={D\bibinitperiod}, + given={Andrew\bibnamedelima J.}, + giveni={A\bibinitperiod\bibinitdelim J\bibinitperiod}}}% + } + \strng{namehash}{cb9d7ae96095d94c44dc40ec02690725} + \strng{fullhash}{e14299d26e49166e6d56990149072c8d} + \strng{bibnamehash}{e14299d26e49166e6d56990149072c8d} + \strng{authorbibnamehash}{e14299d26e49166e6d56990149072c8d} + \strng{authornamehash}{cb9d7ae96095d94c44dc40ec02690725} + \strng{authorfullhash}{e14299d26e49166e6d56990149072c8d} + \field{sortinit}{7} + \field{sortinithash}{108d0be1b1bee9773a1173443802c0a3} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{booktitle}{Proceedings of the {{IEEE}}/{{CVF}} International Conference on Computer Vision} + \field{shorttitle}{Imap} + \field{title}{Imap: {{Implicit}} Mapping and Positioning in Real-Time} + \field{urlday}{10} + \field{urlmonth}{8} + \field{urlyear}{2024} + \field{year}{2021} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{6229\bibrangedash 6238} + \range{pages}{10} + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\experiments\Sucar et al_2021_imap.pdf + \endverb + \verb{urlraw} + \verb http://openaccess.thecvf.com/content/ICCV2021/html/Sucar_iMAP_Implicit_Mapping_and_Positioning_in_Real-Time_ICCV_2021_paper.html + \endverb + \verb{url} + \verb http://openaccess.thecvf.com/content/ICCV2021/html/Sucar_iMAP_Implicit_Mapping_and_Positioning_in_Real-Time_ICCV_2021_paper.html + \endverb + \keyw{⛔ No DOI found} + \endentry + \entry{zhuNiceslamNeuralImplicit2022}{inproceedings}{} + \name{author}{8}{}{% + {{hash=33ed4efed73bba931ef2327c69d56723}{% + family={Zhu}, + familyi={Z\bibinitperiod}, + given={Zihan}, + giveni={Z\bibinitperiod}}}% + {{hash=50cf323c42cc810f490f2ba9cbf452c0}{% + family={Peng}, + familyi={P\bibinitperiod}, + given={Songyou}, + giveni={S\bibinitperiod}}}% + {{hash=594abed54a24fae71414120d58ff9b7f}{% + family={Larsson}, + familyi={L\bibinitperiod}, + given={Viktor}, + giveni={V\bibinitperiod}}}% + {{hash=ff29ced796ff976c25650183cd141ad2}{% + family={Xu}, + familyi={X\bibinitperiod}, + given={Weiwei}, + giveni={W\bibinitperiod}}}% + {{hash=c384a68d4495c23c2753e5bbbad04be0}{% + family={Bao}, + familyi={B\bibinitperiod}, + given={Hujun}, + giveni={H\bibinitperiod}}}% + {{hash=152d039c26f4cbd157947af9ab63a949}{% + family={Cui}, + familyi={C\bibinitperiod}, + given={Zhaopeng}, + giveni={Z\bibinitperiod}}}% + {{hash=44e1d63adacaa62a8b257c28eb6145e2}{% + family={Oswald}, + familyi={O\bibinitperiod}, + given={Martin\bibnamedelima R.}, + giveni={M\bibinitperiod\bibinitdelim R\bibinitperiod}}}% + {{hash=501629493a5ead2d23ee597103478ce8}{% + family={Pollefeys}, + familyi={P\bibinitperiod}, + given={Marc}, + giveni={M\bibinitperiod}}}% + } + \strng{namehash}{23ead790ee2eaa7ca65cdb95443007fd} + \strng{fullhash}{72d20ee339f6a1a73be15d681a8dda6b} + \strng{bibnamehash}{23ead790ee2eaa7ca65cdb95443007fd} + \strng{authorbibnamehash}{23ead790ee2eaa7ca65cdb95443007fd} + \strng{authornamehash}{23ead790ee2eaa7ca65cdb95443007fd} + \strng{authorfullhash}{72d20ee339f6a1a73be15d681a8dda6b} + \field{sortinit}{7} + \field{sortinithash}{108d0be1b1bee9773a1173443802c0a3} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{booktitle}{Proceedings of the {{IEEE}}/{{CVF}} Conference on Computer Vision and Pattern Recognition} + \field{shorttitle}{Nice-Slam} + \field{title}{Nice-Slam: {{Neural}} Implicit Scalable Encoding for Slam} + \field{urlday}{3} + \field{urlmonth}{9} + \field{urlyear}{2024} + \field{year}{2022} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{12786\bibrangedash 12796} + \range{pages}{11} + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\relate_works\Zhu et al_2022_Nice-slam.pdf + \endverb + \verb{urlraw} + \verb http://openaccess.thecvf.com/content/CVPR2022/html/Zhu_NICE-SLAM_Neural_Implicit_Scalable_Encoding_for_SLAM_CVPR_2022_paper.html + \endverb + \verb{url} + \verb http://openaccess.thecvf.com/content/CVPR2022/html/Zhu_NICE-SLAM_Neural_Implicit_Scalable_Encoding_for_SLAM_CVPR_2022_paper.html + \endverb + \keyw{⛔ No DOI found} + \endentry + \entry{garbinFastnerfHighfidelityNeural2021}{inproceedings}{} + \name{author}{5}{}{% + {{hash=96e894055172e371846089961c792631}{% + family={Garbin}, + familyi={G\bibinitperiod}, + given={Stephan\bibnamedelima J.}, + giveni={S\bibinitperiod\bibinitdelim J\bibinitperiod}}}% + {{hash=86a4a43e222898ed1445d733259f6176}{% + family={Kowalski}, + familyi={K\bibinitperiod}, + given={Marek}, + giveni={M\bibinitperiod}}}% + {{hash=2ae3260bfecb5be8445e17eccabf2143}{% + family={Johnson}, + familyi={J\bibinitperiod}, + given={Matthew}, + giveni={M\bibinitperiod}}}% + {{hash=3b90a1e3dad54d8cb1230eec501fdf04}{% + family={Shotton}, + familyi={S\bibinitperiod}, + given={Jamie}, + giveni={J\bibinitperiod}}}% + {{hash=3a5f1e9a85bac703b518103ff99d6300}{% + family={Valentin}, + familyi={V\bibinitperiod}, + given={Julien}, + giveni={J\bibinitperiod}}}% + } + \strng{namehash}{4003902c26a379d8c0759c7897d1733f} + \strng{fullhash}{68bfcfb43a859500f1d4a0ae6cdc75eb} + \strng{bibnamehash}{68bfcfb43a859500f1d4a0ae6cdc75eb} + \strng{authorbibnamehash}{68bfcfb43a859500f1d4a0ae6cdc75eb} + \strng{authornamehash}{4003902c26a379d8c0759c7897d1733f} + \strng{authorfullhash}{68bfcfb43a859500f1d4a0ae6cdc75eb} + \field{sortinit}{8} + \field{sortinithash}{a231b008ebf0ecbe0b4d96dcc159445f} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{booktitle}{Proceedings of the {{IEEE}}/{{CVF}} International Conference on Computer Vision} + \field{shorttitle}{Fastnerf} + \field{title}{Fastnerf: {{High-fidelity}} Neural Rendering at 200fps} + \field{urlday}{26} + \field{urlmonth}{9} + \field{urlyear}{2024} + \field{year}{2021} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{14346\bibrangedash 14355} + \range{pages}{10} + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\relate_works\Garbin et al_2021_Fastnerf.pdf + \endverb + \verb{urlraw} + \verb http://openaccess.thecvf.com/content/ICCV2021/html/Garbin_FastNeRF_High-Fidelity_Neural_Rendering_at_200FPS_ICCV_2021_paper.html + \endverb + \verb{url} + \verb http://openaccess.thecvf.com/content/ICCV2021/html/Garbin_FastNeRF_High-Fidelity_Neural_Rendering_at_200FPS_ICCV_2021_paper.html + \endverb + \keyw{⛔ No DOI found} + \endentry + \entry{kerbl3DGaussianSplatting2023}{article}{} + \name{author}{4}{}{% + {{hash=c73d638ffd25328757379835747302ff}{% + family={Kerbl}, + familyi={K\bibinitperiod}, + given={Bernhard}, + giveni={B\bibinitperiod}}}% + {{hash=2ab758a425dcd6ed440f4ae0d994dc2c}{% + family={Kopanas}, + familyi={K\bibinitperiod}, + given={Georgios}, + giveni={G\bibinitperiod}}}% + {{hash=23f06719145b1e18e13a7e436bb8c2fc}{% + family={Leimkühler}, + familyi={L\bibinitperiod}, + given={Thomas}, + giveni={T\bibinitperiod}}}% + {{hash=70ecb4c4a672f7c20e4b53b57d5a437e}{% + family={Drettakis}, + familyi={D\bibinitperiod}, + given={George}, + giveni={G\bibinitperiod}}}% + } + \list{publisher}{1}{% + {ACM}% + } + \strng{namehash}{38ed96775a79b080a25e1e45e0aaa88a} + \strng{fullhash}{36d84a8a291db8ba20b3f78b93a2eeee} + \strng{bibnamehash}{36d84a8a291db8ba20b3f78b93a2eeee} + \strng{authorbibnamehash}{36d84a8a291db8ba20b3f78b93a2eeee} + \strng{authornamehash}{38ed96775a79b080a25e1e45e0aaa88a} + \strng{authorfullhash}{36d84a8a291db8ba20b3f78b93a2eeee} + \field{sortinit}{9} + \field{sortinithash}{0a5ebc79d83c96b6579069544c73c7d4} + \field{labelnamesource}{author} + \field{labeltitlesource}{title} + \field{journaltitle}{ACM Transactions on Graphics} + \field{number}{4} + \field{title}{{{3D Gaussian Splatting}} for {{Real-Time Radiance Field Rendering}}} + \field{urlday}{9} + \field{urlmonth}{6} + \field{urlyear}{2024} + \field{volume}{42} + \field{year}{2023} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{1\bibrangedash 14} + \range{pages}{14} + \verb{doi} + \verb 10.1145/3592433 + \endverb + \verb{file} + \verb C\:\\Users\\18317\\DevSpace\\DocHub\\essay\\zotero\\attachments\\gs_spalt\\Kerbl et al_2023_3d gaussian splatting for real-time radiance field rendering.pdf;C\:\\Users\\18317\\Zotero\\storage\\PRZ6GF98\\3D Gaussian Splatting for Real-Time Radiance Field Rendering - 3d_gaussian_splatting_low.pdf + \endverb + \verb{urlraw} + \verb https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/ + \endverb + \verb{url} + \verb https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/ + \endverb + \endentry + \entry{pengRTGSLAMRealtime3D2024}{online}{} + \name{author}{7}{}{% + {{hash=913cd183a9be76d37247a590b24fd552}{% + family={Peng}, + familyi={P\bibinitperiod}, + given={Zhexi}, + giveni={Z\bibinitperiod}}}% + {{hash=9228bb2602666eea3f3f7cec2d5e8c34}{% + family={Shao}, + familyi={S\bibinitperiod}, + given={Tianjia}, + giveni={T\bibinitperiod}}}% + {{hash=bc6c6f66f139beef784a50d9f1485019}{% + family={Liu}, + familyi={L\bibinitperiod}, + given={Yong}, + giveni={Y\bibinitperiod}}}% + {{hash=0e1bc1d9a4e5ea73ad61486fb9b85ff6}{% + family={Zhou}, + familyi={Z\bibinitperiod}, + given={Jingke}, + giveni={J\bibinitperiod}}}% + {{hash=dc9616dbd138c8816662e7a1bf22bc60}{% + family={Yang}, + familyi={Y\bibinitperiod}, + given={Yin}, + giveni={Y\bibinitperiod}}}% + {{hash=06b47d0dcf324aeafa35d1a8666637ba}{% + family={Wang}, + familyi={W\bibinitperiod}, + given={Jingdong}, + giveni={J\bibinitperiod}}}% + {{hash=09d5d500b34786c94a128e6f131300c6}{% + family={Zhou}, + familyi={Z\bibinitperiod}, + given={Kun}, + giveni={K\bibinitperiod}}}% + } + \strng{namehash}{e1660083991b1325b562914dabe293c0} + \strng{fullhash}{4e1c25d3d8e961a2e5c238289b53fa8b} + \strng{bibnamehash}{e1660083991b1325b562914dabe293c0} + \strng{authorbibnamehash}{e1660083991b1325b562914dabe293c0} + \strng{authornamehash}{e1660083991b1325b562914dabe293c0} + \strng{authorfullhash}{4e1c25d3d8e961a2e5c238289b53fa8b} + \field{sortinit}{1} + \field{sortinithash}{4f6aaa89bab872aa0999fec09ff8e98a} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{abstract}{We present Real-time Gaussian SLAM (RTG-SLAM), a real-time 3D reconstruction system with an RGBD camera for large-scale environments using Gaussian splatting. The system features a compact Gaussian representation and a highly efficient on-the-fly Gaussian optimization scheme. We force each Gaussian to be either opaque or nearly transparent, with the opaque ones fitting the surface and dominant colors, and transparent ones fitting residual colors. By rendering depth in a different way from color rendering, we let a single opaque Gaussian well fit a local surface region without the need of multiple overlapping Gaussians, hence largely reducing the memory and computation cost. For on-the-fly Gaussian optimization, we explicitly add Gaussians for three types of pixels per frame: newly observed, with large color errors, and with large depth errors. We also categorize all Gaussians into stable and unstable ones, where the stable Gaussians are expected to well fit previously observed RGBD images and otherwise unstable. We only optimize the unstable Gaussians and only render the pixels occupied by unstable Gaussians. In this way, both the number of Gaussians to be optimized and pixels to be rendered are largely reduced, and the optimization can be done in real time. We show real-time reconstructions of a variety of large scenes. Compared with the state-of-the-art NeRF-based RGBD SLAM, our system achieves comparable high-quality reconstruction but with around twice the speed and half the memory cost, and shows superior performance in the realism of novel view synthesis and camera tracking accuracy.} + \field{day}{8} + \field{eprintclass}{cs} + \field{eprinttype}{arXiv} + \field{month}{5} + \field{pubstate}{prepublished} + \field{shorttitle}{{{RTG-SLAM}}} + \field{title}{{{RTG-SLAM}}: {{Real-time 3D Reconstruction}} at {{Scale}} Using {{Gaussian Splatting}}} + \field{urlday}{16} + \field{urlmonth}{7} + \field{urlyear}{2024} + \field{year}{2024} + \field{dateera}{ce} + \field{urldateera}{ce} + \verb{doi} + \verb 10.1145/3658233 + \endverb + \verb{eprint} + \verb 2404.19706 + \endverb + \verb{file} + \verb C\:\\Users\\18317\\DevSpace\\DocHub\\essay\\zotero\\attachments\\gs_spalt\\Peng et al_2024_RTG-SLAM2.pdf;C\:\\Users\\18317\\Zotero\\storage\\XV5IHMT8\\2404.html + \endverb + \verb{urlraw} + \verb http://arxiv.org/abs/2404.19706 + \endverb + \verb{url} + \verb http://arxiv.org/abs/2404.19706 + \endverb + \keyw{Computer Science - Computer Vision and Pattern Recognition} + \endentry + \entry{haRGBDGSICPSLAM2024}{online}{} + \name{author}{3}{}{% + {{hash=3db90f36d26bf6e32b86a0f10ed792d9}{% + family={Ha}, + familyi={H\bibinitperiod}, + given={Seongbo}, + giveni={S\bibinitperiod}}}% + {{hash=cc16fc0f9a053d60092ca6a16d14dbde}{% + family={Yeon}, + familyi={Y\bibinitperiod}, + given={Jiung}, + giveni={J\bibinitperiod}}}% + {{hash=ff40e96d1533de2d7b3715b675186f41}{% + family={Yu}, + familyi={Y\bibinitperiod}, + given={Hyeonwoo}, + giveni={H\bibinitperiod}}}% + } + \strng{namehash}{9268e354cd202e63d4eba6374dc788c7} + \strng{fullhash}{7b15c6cdf3cbf35f4b9d9a68bc0bb2b6} + \strng{bibnamehash}{7b15c6cdf3cbf35f4b9d9a68bc0bb2b6} + \strng{authorbibnamehash}{7b15c6cdf3cbf35f4b9d9a68bc0bb2b6} + \strng{authornamehash}{9268e354cd202e63d4eba6374dc788c7} + \strng{authorfullhash}{7b15c6cdf3cbf35f4b9d9a68bc0bb2b6} + \field{sortinit}{1} + \field{sortinithash}{4f6aaa89bab872aa0999fec09ff8e98a} + \field{labelnamesource}{author} + \field{labeltitlesource}{title} + \field{abstract}{Simultaneous Localization and Mapping (SLAM) with dense representation plays a key role in robotics, Virtual Reality (VR), and Augmented Reality (AR) applications. Recent advancements in dense representation SLAM have highlighted the potential of leveraging neural scene representation and 3D Gaussian representation for high-fidelity spatial representation. In this paper, we propose a novel dense representation SLAM approach with a fusion of Generalized Iterative Closest Point (G-ICP) and 3D Gaussian Splatting (3DGS). In contrast to existing methods, we utilize a single Gaussian map for both tracking and mapping, resulting in mutual benefits. Through the exchange of covariances between tracking and mapping processes with scale alignment techniques, we minimize redundant computations and achieve an efficient system. Additionally, we enhance tracking accuracy and mapping quality through our keyframe selection methods. Experimental results demonstrate the effectiveness of our approach, showing an incredibly fast speed up to 107 FPS (for the entire system) and superior quality of the reconstructed map.} + \field{day}{22} + \field{eprintclass}{cs} + \field{eprinttype}{arXiv} + \field{month}{3} + \field{pubstate}{prepublished} + \field{title}{{{RGBD GS-ICP SLAM}}} + \field{urlday}{23} + \field{urlmonth}{5} + \field{urlyear}{2024} + \field{year}{2024} + \field{dateera}{ce} + \field{urldateera}{ce} + \verb{eprint} + \verb 2403.12550 + \endverb + \verb{file} + \verb C\:\\Users\\18317\\Zotero\\storage\\HJ4D53M9\\Ha et al. - 2024 - RGBD GS-ICP SLAM.pdf;C\:\\Users\\18317\\Zotero\\storage\\P52BZ3I7\\2403.html + \endverb + \verb{urlraw} + \verb http://arxiv.org/abs/2403.12550 + \endverb + \verb{url} + \verb http://arxiv.org/abs/2403.12550 + \endverb + \keyw{Computer Science - Computer Vision and Pattern Recognition} + \endentry + \entry{yugayGaussianSLAMPhotorealisticDense2024}{online}{} + \name{author}{4}{}{% + {{hash=c0d899a4b2ad76e85b5a10bfa35ad4e7}{% + family={Yugay}, + familyi={Y\bibinitperiod}, + given={Vladimir}, + giveni={V\bibinitperiod}}}% + {{hash=0e1dff109644a2423d611a0bfe123178}{% + family={Li}, + familyi={L\bibinitperiod}, + given={Yue}, + giveni={Y\bibinitperiod}}}% + {{hash=50a45145d1cd6b9c9782801e107d49eb}{% + family={Gevers}, + familyi={G\bibinitperiod}, + given={Theo}, + giveni={T\bibinitperiod}}}% + {{hash=44e1d63adacaa62a8b257c28eb6145e2}{% + family={Oswald}, + familyi={O\bibinitperiod}, + given={Martin\bibnamedelima R.}, + giveni={M\bibinitperiod\bibinitdelim R\bibinitperiod}}}% + } + \strng{namehash}{c8095d8c9b833b7f754889b60070f3a0} + \strng{fullhash}{1e276592282b2ed69ec349557c18d6c1} + \strng{bibnamehash}{1e276592282b2ed69ec349557c18d6c1} + \strng{authorbibnamehash}{1e276592282b2ed69ec349557c18d6c1} + \strng{authornamehash}{c8095d8c9b833b7f754889b60070f3a0} + \strng{authorfullhash}{1e276592282b2ed69ec349557c18d6c1} + \field{sortinit}{1} + \field{sortinithash}{4f6aaa89bab872aa0999fec09ff8e98a} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{abstract}{We present a dense simultaneous localization and mapping (SLAM) method that uses 3D Gaussians as a scene representation. Our approach enables interactive-time reconstruction and photo-realistic rendering from real-world single-camera RGBD videos. To this end, we propose a novel effective strategy for seeding new Gaussians for newly explored areas and their effective online optimization that is independent of the scene size and thus scalable to larger scenes. This is achieved by organizing the scene into sub-maps which are independently optimized and do not need to be kept in memory. We further accomplish frame-to-model camera tracking by minimizing photometric and geometric losses between the input and rendered frames. The Gaussian representation allows for high-quality photo-realistic real-time rendering of real-world scenes. Evaluation on synthetic and real-world datasets demonstrates competitive or superior performance in mapping, tracking, and rendering compared to existing neural dense SLAM methods.} + \field{day}{22} + \field{eprintclass}{cs} + \field{eprinttype}{arXiv} + \field{month}{3} + \field{pubstate}{prepublished} + \field{shorttitle}{Gaussian-{{SLAM}}} + \field{title}{Gaussian-{{SLAM}}: {{Photo-realistic Dense SLAM}} with {{Gaussian Splatting}}} + \field{urlday}{9} + \field{urlmonth}{6} + \field{urlyear}{2024} + \field{year}{2024} + \field{dateera}{ce} + \field{urldateera}{ce} + \verb{eprint} + \verb 2312.10070 + \endverb + \verb{file} + \verb C\:\\Users\\18317\\DevSpace\\DocHub\\essay\\zotero\\attachments\\gs_spalt\\Yugay et al_2024_Gaussian-SLAM2.pdf;C\:\\Users\\18317\\Zotero\\storage\\9J38D593\\2312.html + \endverb + \verb{urlraw} + \verb http://arxiv.org/abs/2312.10070 + \endverb + \verb{url} + \verb http://arxiv.org/abs/2312.10070 + \endverb + \keyw{Computer Science - Computer Vision and Pattern Recognition,Computer Science - Robotics} + \endentry + \entry{mur-artalOrbslam2OpensourceSlam2017}{article}{} + \name{author}{2}{}{% + {{hash=3bcf9140fcfbbb3fa5cdedbcb3380ee8}{% + family={Mur-Artal}, + familyi={M\bibinithyphendelim A\bibinitperiod}, + given={Raul}, + giveni={R\bibinitperiod}}}% + {{hash=f309368530c6732a6657d7380a443890}{% + family={Tardós}, + familyi={T\bibinitperiod}, + given={Juan\bibnamedelima D.}, + giveni={J\bibinitperiod\bibinitdelim D\bibinitperiod}}}% + } + \list{publisher}{1}{% + {IEEE}% + } + \strng{namehash}{97ed3709bd3c4e34f07fd331ed1bb6b8} + \strng{fullhash}{97ed3709bd3c4e34f07fd331ed1bb6b8} + \strng{bibnamehash}{97ed3709bd3c4e34f07fd331ed1bb6b8} + \strng{authorbibnamehash}{97ed3709bd3c4e34f07fd331ed1bb6b8} + \strng{authornamehash}{97ed3709bd3c4e34f07fd331ed1bb6b8} + \strng{authorfullhash}{97ed3709bd3c4e34f07fd331ed1bb6b8} + \field{sortinit}{1} + \field{sortinithash}{4f6aaa89bab872aa0999fec09ff8e98a} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{journaltitle}{IEEE transactions on robotics} + \field{number}{5} + \field{shorttitle}{Orb-Slam2} + \field{title}{Orb-Slam2: {{An}} Open-Source Slam System for Monocular, Stereo, and Rgb-d Cameras} + \field{urlday}{22} + \field{urlmonth}{8} + \field{urlyear}{2024} + \field{volume}{33} + \field{year}{2017} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{1255\bibrangedash 1262} + \range{pages}{8} + \verb{doi} + \verb 10.1109/TRO.2017.2705103 + \endverb + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\experiments\Mur-Artal_Tardós_2017_Orb-slam2.pdf + \endverb + \verb{urlraw} + \verb https://ieeexplore.ieee.org/abstract/document/7946260/ + \endverb + \verb{url} + \verb https://ieeexplore.ieee.org/abstract/document/7946260/ + \endverb + \endentry + \entry{camposOrbslam3AccurateOpensource2021}{article}{} + \name{author}{5}{}{% + {{hash=6cdacf84be87337589747b5b4820bfba}{% + family={Campos}, + familyi={C\bibinitperiod}, + given={Carlos}, + giveni={C\bibinitperiod}}}% + {{hash=45619e807fa5241f5da5135ce7992309}{% + family={Elvira}, + familyi={E\bibinitperiod}, + given={Richard}, + giveni={R\bibinitperiod}}}% + {{hash=9f43a0aba53c88080ad5450b521f709d}{% + family={Rodríguez}, + familyi={R\bibinitperiod}, + given={Juan\bibnamedelimb J.\bibnamedelimi Gómez}, + giveni={J\bibinitperiod\bibinitdelim J\bibinitperiod\bibinitdelim G\bibinitperiod}}}% + {{hash=96f8ef4e5aa02ab674ae4c12e943e56b}{% + family={Montiel}, + familyi={M\bibinitperiod}, + given={José\bibnamedelima MM}, + giveni={J\bibinitperiod\bibinitdelim M\bibinitperiod}}}% + {{hash=f309368530c6732a6657d7380a443890}{% + family={Tardós}, + familyi={T\bibinitperiod}, + given={Juan\bibnamedelima D.}, + giveni={J\bibinitperiod\bibinitdelim D\bibinitperiod}}}% + } + \list{publisher}{1}{% + {IEEE}% + } + \strng{namehash}{4b5f61bd95eca9030bea99344bcf2557} + \strng{fullhash}{119a564e6f05f80e0dea9257eb4dd906} + \strng{bibnamehash}{119a564e6f05f80e0dea9257eb4dd906} + \strng{authorbibnamehash}{119a564e6f05f80e0dea9257eb4dd906} + \strng{authornamehash}{4b5f61bd95eca9030bea99344bcf2557} + \strng{authorfullhash}{119a564e6f05f80e0dea9257eb4dd906} + \field{sortinit}{1} + \field{sortinithash}{4f6aaa89bab872aa0999fec09ff8e98a} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{journaltitle}{IEEE Transactions on Robotics} + \field{number}{6} + \field{shorttitle}{Orb-Slam3} + \field{title}{Orb-Slam3: {{An}} Accurate Open-Source Library for Visual, Visual–Inertial, and Multimap Slam} + \field{urlday}{20} + \field{urlmonth}{8} + \field{urlyear}{2024} + \field{volume}{37} + \field{year}{2021} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{1874\bibrangedash 1890} + \range{pages}{17} + \verb{doi} + \verb 10.1109/TRO.2021.3075644 + \endverb + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\experiments\Campos et al_2021_Orb-slam3.pdf + \endverb + \verb{urlraw} + \verb https://ieeexplore.ieee.org/abstract/document/9440682/ + \endverb + \verb{url} + \verb https://ieeexplore.ieee.org/abstract/document/9440682/ + \endverb + \endentry + \entry{gauglitzEvaluationInterestPoint2011}{article}{} + \name{author}{3}{}{% + {{hash=d495c6381639662493aa21ba580c2d83}{% + family={Gauglitz}, + familyi={G\bibinitperiod}, + given={Steffen}, + giveni={S\bibinitperiod}}}% + {{hash=22f545ecadaedf548ce1942093a0a225}{% + family={Höllerer}, + familyi={H\bibinitperiod}, + given={Tobias}, + giveni={T\bibinitperiod}}}% + {{hash=3688cccd8484f3ca2424096f189fed1c}{% + family={Turk}, + familyi={T\bibinitperiod}, + given={Matthew}, + giveni={M\bibinitperiod}}}% + } + \strng{namehash}{88d978d7c5a4cc39e9abf29f5fb26a04} + \strng{fullhash}{7e76681aa49be53850861f6b3e19af63} + \strng{bibnamehash}{7e76681aa49be53850861f6b3e19af63} + \strng{authorbibnamehash}{7e76681aa49be53850861f6b3e19af63} + \strng{authornamehash}{88d978d7c5a4cc39e9abf29f5fb26a04} + \strng{authorfullhash}{7e76681aa49be53850861f6b3e19af63} + \field{sortinit}{1} + \field{sortinithash}{4f6aaa89bab872aa0999fec09ff8e98a} + \field{labelnamesource}{author} + \field{labeltitlesource}{title} + \field{issn}{0920-5691, 1573-1405} + \field{journaltitle}{International Journal of Computer Vision} + \field{langid}{english} + \field{month}{9} + \field{number}{3} + \field{shortjournal}{Int J Comput Vis} + \field{title}{Evaluation of {{Interest Point Detectors}} and {{Feature Descriptors}} for {{Visual Tracking}}} + \field{urlday}{26} + \field{urlmonth}{9} + \field{urlyear}{2024} + \field{volume}{94} + \field{year}{2011} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{335\bibrangedash 360} + \range{pages}{26} + \verb{doi} + \verb 10.1007/s11263-011-0431-5 + \endverb + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\relate_works\Gauglitz et al_2011_Evaluation of Interest Point Detectors and Feature Descriptors for Visual.pdf + \endverb + \verb{urlraw} + \verb http://link.springer.com/10.1007/s11263-011-0431-5 + \endverb + \verb{url} + \verb http://link.springer.com/10.1007/s11263-011-0431-5 + \endverb + \endentry + \entry{engelDirectSparseOdometry2017}{article}{} + \name{author}{3}{}{% + {{hash=105f32574a19425c80138198b0b09cd5}{% + family={Engel}, + familyi={E\bibinitperiod}, + given={Jakob}, + giveni={J\bibinitperiod}}}% + {{hash=9025300f43a193cfed8e10e924f2a117}{% + family={Koltun}, + familyi={K\bibinitperiod}, + given={Vladlen}, + giveni={V\bibinitperiod}}}% + {{hash=1bd2b6b6ca2fc15a90f164070b626131}{% + family={Cremers}, + familyi={C\bibinitperiod}, + given={Daniel}, + giveni={D\bibinitperiod}}}% + } + \list{publisher}{1}{% + {IEEE}% + } + \strng{namehash}{3679e30c905fef5f09a6d8d4a06f62fc} + \strng{fullhash}{974ceb49e0d7e4e5dec3dc9cf0eaeae5} + \strng{bibnamehash}{974ceb49e0d7e4e5dec3dc9cf0eaeae5} + \strng{authorbibnamehash}{974ceb49e0d7e4e5dec3dc9cf0eaeae5} + \strng{authornamehash}{3679e30c905fef5f09a6d8d4a06f62fc} + \strng{authorfullhash}{974ceb49e0d7e4e5dec3dc9cf0eaeae5} + \field{sortinit}{1} + \field{sortinithash}{4f6aaa89bab872aa0999fec09ff8e98a} + \field{labelnamesource}{author} + \field{labeltitlesource}{title} + \field{journaltitle}{IEEE transactions on pattern analysis and machine intelligence} + \field{number}{3} + \field{title}{Direct Sparse Odometry} + \field{urlday}{3} + \field{urlmonth}{9} + \field{urlyear}{2024} + \field{volume}{40} + \field{year}{2017} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{611\bibrangedash 625} + \range{pages}{15} + \verb{doi} + \verb 10.1109/TPAMI.2017.2658577 + \endverb + \verb{urlraw} + \verb https://ieeexplore.ieee.org/abstract/document/7898369/ + \endverb + \verb{url} + \verb https://ieeexplore.ieee.org/abstract/document/7898369/ + \endverb + \keyw{nosource} + \endentry + \entry{kerlRobustOdometryEstimation2013}{inproceedings}{} + \name{author}{3}{}{% + {{hash=c8b46d457d7577efc9d388d1c40fb783}{% + family={Kerl}, + familyi={K\bibinitperiod}, + given={Christian}, + giveni={C\bibinitperiod}}}% + {{hash=9e291584a3b29b27c88ecfe0a566274c}{% + family={Sturm}, + familyi={S\bibinitperiod}, + given={Jürgen}, + giveni={J\bibinitperiod}}}% + {{hash=1bd2b6b6ca2fc15a90f164070b626131}{% + family={Cremers}, + familyi={C\bibinitperiod}, + given={Daniel}, + giveni={D\bibinitperiod}}}% + } + \list{publisher}{1}{% + {IEEE}% + } + \strng{namehash}{4088c2282796db3475bd1d361a9ac3f1} + \strng{fullhash}{21c79321a177b2edd1d9ad2ca8f11e28} + \strng{bibnamehash}{21c79321a177b2edd1d9ad2ca8f11e28} + \strng{authorbibnamehash}{21c79321a177b2edd1d9ad2ca8f11e28} + \strng{authornamehash}{4088c2282796db3475bd1d361a9ac3f1} + \strng{authorfullhash}{21c79321a177b2edd1d9ad2ca8f11e28} + \field{extraname}{2} + \field{sortinit}{1} + \field{sortinithash}{4f6aaa89bab872aa0999fec09ff8e98a} + \field{labelnamesource}{author} + \field{labeltitlesource}{title} + \field{booktitle}{2013 {{IEEE}} International Conference on Robotics and Automation} + \field{title}{Robust Odometry Estimation for {{RGB-D}} Cameras} + \field{urlday}{13} + \field{urlmonth}{9} + \field{urlyear}{2024} + \field{year}{2013} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{3748\bibrangedash 3754} + \range{pages}{7} + \verb{doi} + \verb 10.1109/ICRA.2013.6631104 + \endverb + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\relate_works\Kerl et al_2013_Robust odometry estimation for RGB-D cameras.pdf + \endverb + \verb{urlraw} + \verb https://ieeexplore.ieee.org/abstract/document/6631104/ + \endverb + \verb{url} + \verb https://ieeexplore.ieee.org/abstract/document/6631104/ + \endverb + \endentry + \entry{newcombeDTAMDenseTracking2011}{inproceedings}{} + \name{author}{3}{}{% + {{hash=7b635dacb5ef25a6f9c500f8c62aacd8}{% + family={Newcombe}, + familyi={N\bibinitperiod}, + given={Richard\bibnamedelima A.}, + giveni={R\bibinitperiod\bibinitdelim A\bibinitperiod}}}% + {{hash=2900c8a8d13f9001d8cd329cee1bb8da}{% + family={Lovegrove}, + familyi={L\bibinitperiod}, + given={Steven\bibnamedelima J.}, + giveni={S\bibinitperiod\bibinitdelim J\bibinitperiod}}}% + {{hash=6800da0b1eba1591ea0c413f8efceae3}{% + family={Davison}, + familyi={D\bibinitperiod}, + given={Andrew\bibnamedelima J.}, + giveni={A\bibinitperiod\bibinitdelim J\bibinitperiod}}}% + } + \list{publisher}{1}{% + {IEEE}% + } + \strng{namehash}{1716b787f56b1649cb7cef04aa0bc5c4} + \strng{fullhash}{64b31b2d18f97261375befd7160afdf7} + \strng{bibnamehash}{64b31b2d18f97261375befd7160afdf7} + \strng{authorbibnamehash}{64b31b2d18f97261375befd7160afdf7} + \strng{authornamehash}{1716b787f56b1649cb7cef04aa0bc5c4} + \strng{authorfullhash}{64b31b2d18f97261375befd7160afdf7} + \field{extraname}{2} + \field{sortinit}{1} + \field{sortinithash}{4f6aaa89bab872aa0999fec09ff8e98a} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{booktitle}{2011 International Conference on Computer Vision} + \field{shorttitle}{{{DTAM}}} + \field{title}{{{DTAM}}: {{Dense}} Tracking and Mapping in Real-Time} + \field{urlday}{13} + \field{urlmonth}{9} + \field{urlyear}{2024} + \field{year}{2011} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{2320\bibrangedash 2327} + \range{pages}{8} + \verb{doi} + \verb 10.1109/ICCV.2011.6126513 + \endverb + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\relate_works\Newcombe et al_2011_DTAM.pdf + \endverb + \verb{urlraw} + \verb https://ieeexplore.ieee.org/abstract/document/6126513/ + \endverb + \verb{url} + \verb https://ieeexplore.ieee.org/abstract/document/6126513/ + \endverb + \endentry + \entry{whelanElasticFusionRealtimeDense2016}{article}{} + \name{author}{5}{}{% + {{hash=547364b3c154f5f2f7d53a4542e8e7c4}{% + family={Whelan}, + familyi={W\bibinitperiod}, + given={Thomas}, + giveni={T\bibinitperiod}}}% + {{hash=7cd6e12dc02cf694c085b89c2a35d7a8}{% + family={Salas-Moreno}, + familyi={S\bibinithyphendelim M\bibinitperiod}, + given={Renato\bibnamedelima F}, + giveni={R\bibinitperiod\bibinitdelim F\bibinitperiod}}}% + {{hash=5bc19448e90c8f3e88e5fbd50f12ac15}{% + family={Glocker}, + familyi={G\bibinitperiod}, + given={Ben}, + giveni={B\bibinitperiod}}}% + {{hash=6800da0b1eba1591ea0c413f8efceae3}{% + family={Davison}, + familyi={D\bibinitperiod}, + given={Andrew\bibnamedelima J}, + giveni={A\bibinitperiod\bibinitdelim J\bibinitperiod}}}% + {{hash=25564f7f36f46be27f7326210cc5c2ad}{% + family={Leutenegger}, + familyi={L\bibinitperiod}, + given={Stefan}, + giveni={S\bibinitperiod}}}% + } + \strng{namehash}{9809af5f93be49cc52c15121e141a753} + \strng{fullhash}{f5cd21bb0933edf7c64b11c1baf3e2c7} + \strng{bibnamehash}{f5cd21bb0933edf7c64b11c1baf3e2c7} + \strng{authorbibnamehash}{f5cd21bb0933edf7c64b11c1baf3e2c7} + \strng{authornamehash}{9809af5f93be49cc52c15121e141a753} + \strng{authorfullhash}{f5cd21bb0933edf7c64b11c1baf3e2c7} + \field{sortinit}{1} + \field{sortinithash}{4f6aaa89bab872aa0999fec09ff8e98a} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{abstract}{We present a novel approach to real-time dense visual simultaneous localisation and mapping. Our system is capable of capturing comprehensive dense globally consistent surfel-based maps of room scale environments and beyond explored using an RGB-D camera in an incremental online fashion, without pose graph optimization or any post-processing steps. This is accomplished by using dense frame-to-model camera tracking and windowed surfel-based fusion coupled with frequent model refinement through non-rigid surface deformations. Our approach applies local model-to-model surface loop closure optimizations as often as possible to stay close to the mode of the map distribution, while utilizing global loop closure to recover from arbitrary drift and maintain global consistency. In the spirit of improving map quality as well as tracking accuracy and robustness, we furthermore explore a novel approach to real-time discrete light source detection. This technique is capable of detecting numerous light sources in indoor environments in real-time as a user handheld camera explores the scene. Absolutely no prior information about the scene or number of light sources is required. By making a small set of simple assumptions about the appearance properties of the scene our method can incrementally estimate both the quantity and location of multiple light sources in the environment in an online fashion. Our results demonstrate that our technique functions well in many different environments and lighting configurations. We show that this enables (a) more realistic augmented reality rendering; (b) a richer understanding of the scene beyond pure geometry and; (c) more accurate and robust photometric tracking.} + \field{issn}{0278-3649, 1741-3176} + \field{journaltitle}{The International Journal of Robotics Research} + \field{langid}{english} + \field{month}{12} + \field{number}{14} + \field{shortjournal}{The International Journal of Robotics Research} + \field{shorttitle}{{{ElasticFusion}}} + \field{title}{{{ElasticFusion}}: {{Real-time}} Dense {{SLAM}} and Light Source Estimation} + \field{urlday}{13} + \field{urlmonth}{9} + \field{urlyear}{2024} + \field{volume}{35} + \field{year}{2016} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{1697\bibrangedash 1716} + \range{pages}{20} + \verb{doi} + \verb 10.1177/0278364916669237 + \endverb + \verb{file} + \verb C\:\\Users\\18317\\DevSpace\\DocHub\\essay\\zotero\\attachments\\gs_spalt\\relate_works\\Whelan et al_2016_ElasticFusion.pdf;C\:\\Users\\18317\\DevSpace\\DocHub\\essay\\zotero\\attachments\\gs_spalt\\relate_works\\Whelan et al_2016_ElasticFusion2.pdf + \endverb + \verb{urlraw} + \verb http://journals.sagepub.com/doi/10.1177/0278364916669237 + \endverb + \verb{url} + \verb http://journals.sagepub.com/doi/10.1177/0278364916669237 + \endverb + \endentry + \entry{yen-chenInerfInvertingNeural2021}{inproceedings}{} + \name{author}{6}{}{% + {{hash=8da2c855845ec964a310dda43392127d}{% + family={Yen-Chen}, + familyi={Y\bibinithyphendelim C\bibinitperiod}, + given={Lin}, + giveni={L\bibinitperiod}}}% + {{hash=8b9d4cb970dc3176b51235d9a18dce48}{% + family={Florence}, + familyi={F\bibinitperiod}, + given={Pete}, + giveni={P\bibinitperiod}}}% + {{hash=562f3254a241932db4a6d44a97473ec5}{% + family={Barron}, + familyi={B\bibinitperiod}, + given={Jonathan\bibnamedelima T.}, + giveni={J\bibinitperiod\bibinitdelim T\bibinitperiod}}}% + {{hash=51675f0d612799bd99aab6d0b9a38088}{% + family={Rodriguez}, + familyi={R\bibinitperiod}, + given={Alberto}, + giveni={A\bibinitperiod}}}% + {{hash=cae9f806bc99a5f19fadea538fc2db04}{% + family={Isola}, + familyi={I\bibinitperiod}, + given={Phillip}, + giveni={P\bibinitperiod}}}% + {{hash=08f925fe4692d130a1d7cb7d94483351}{% + family={Lin}, + familyi={L\bibinitperiod}, + given={Tsung-Yi}, + giveni={T\bibinithyphendelim Y\bibinitperiod}}}% + } + \list{publisher}{1}{% + {IEEE}% + } + \strng{namehash}{699d21985071744ba8b4323771dd610e} + \strng{fullhash}{3e8d26e295acb2d47b734d692fe52a3a} + \strng{bibnamehash}{3e8d26e295acb2d47b734d692fe52a3a} + \strng{authorbibnamehash}{3e8d26e295acb2d47b734d692fe52a3a} + \strng{authornamehash}{699d21985071744ba8b4323771dd610e} + \strng{authorfullhash}{3e8d26e295acb2d47b734d692fe52a3a} + \field{sortinit}{2} + \field{sortinithash}{8b555b3791beccb63322c22f3320aa9a} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{booktitle}{2021 {{IEEE}}/{{RSJ International Conference}} on {{Intelligent Robots}} and {{Systems}} ({{IROS}})} + \field{shorttitle}{Inerf} + \field{title}{Inerf: {{Inverting}} Neural Radiance Fields for Pose Estimation} + \field{urlday}{13} + \field{urlmonth}{9} + \field{urlyear}{2024} + \field{year}{2021} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{1323\bibrangedash 1330} + \range{pages}{8} + \verb{doi} + \verb 10.1109/IROS51168.2021.9636708 + \endverb + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\relate_works\Yen-Chen et al_2021_inerf.pdf + \endverb + \verb{urlraw} + \verb https://ieeexplore.ieee.org/abstract/document/9636708/ + \endverb + \verb{url} + \verb https://ieeexplore.ieee.org/abstract/document/9636708/ + \endverb + \endentry + \entry{mullerInstantNeuralGraphics2022}{article}{} + \name{author}{4}{}{% + {{hash=c2ad285eb682bccdfe2685694db2188f}{% + family={Müller}, + familyi={M\bibinitperiod}, + given={Thomas}, + giveni={T\bibinitperiod}}}% + {{hash=384420ee5132168a95e9f097058e3c55}{% + family={Evans}, + familyi={E\bibinitperiod}, + given={Alex}, + giveni={A\bibinitperiod}}}% + {{hash=56d271cf4aa34e6b3942487f50bc0419}{% + family={Schied}, + familyi={S\bibinitperiod}, + given={Christoph}, + giveni={C\bibinitperiod}}}% + {{hash=611ada91d1e542d341733a32d3b74766}{% + family={Keller}, + familyi={K\bibinitperiod}, + given={Alexander}, + giveni={A\bibinitperiod}}}% + } + \strng{namehash}{ac68b03d9bae4f1b128311bafa44418d} + \strng{fullhash}{5ee9a8053cd04c40f99a57f291d2c4a8} + \strng{bibnamehash}{5ee9a8053cd04c40f99a57f291d2c4a8} + \strng{authorbibnamehash}{5ee9a8053cd04c40f99a57f291d2c4a8} + \strng{authornamehash}{ac68b03d9bae4f1b128311bafa44418d} + \strng{authorfullhash}{5ee9a8053cd04c40f99a57f291d2c4a8} + \field{sortinit}{2} + \field{sortinithash}{8b555b3791beccb63322c22f3320aa9a} + \field{labelnamesource}{author} + \field{labeltitlesource}{title} + \field{abstract}{Neural graphics primitives, parameterized by fully connected neural networks, can be costly to train and evaluate. We reduce this cost with a versatile new input encoding that permits the use of a smaller network without sacrificing quality, thus significantly reducing the number of floating point and memory access operations: a small neural network is augmented by a multiresolution hash table of trainable feature vectors whose values are optimized through stochastic gradient descent. The multiresolution structure allows the network to disambiguate hash collisions, making for a simple architecture that is trivial to parallelize on modern GPUs. We leverage this parallelism by implementing the whole system using fully-fused CUDA kernels with a focus on minimizing wasted bandwidth and compute operations. We achieve a combined speedup of several orders of magnitude, enabling training of high-quality neural graphics primitives in a matter of seconds, and rendering in tens of milliseconds at a resolution of 1920×1080.} + \field{issn}{0730-0301, 1557-7368} + \field{journaltitle}{ACM Transactions on Graphics} + \field{langid}{english} + \field{month}{7} + \field{number}{4} + \field{shortjournal}{ACM Trans. Graph.} + \field{title}{Instant Neural Graphics Primitives with a Multiresolution Hash Encoding} + \field{urlday}{3} + \field{urlmonth}{9} + \field{urlyear}{2024} + \field{volume}{41} + \field{year}{2022} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{1\bibrangedash 15} + \range{pages}{15} + \verb{doi} + \verb 10.1145/3528223.3530127 + \endverb + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\relate_works\Müller et al_2022_Instant neural graphics primitives with a multiresolution hash encoding.pdf + \endverb + \verb{urlraw} + \verb https://dl.acm.org/doi/10.1145/3528223.3530127 + \endverb + \verb{url} + \verb https://dl.acm.org/doi/10.1145/3528223.3530127 + \endverb + \endentry + \entry{yuPlenoctreesRealtimeRendering2021}{inproceedings}{} + \name{author}{6}{}{% + {{hash=ff32d51d3d43688459e059d03b115eca}{% + family={Yu}, + familyi={Y\bibinitperiod}, + given={Alex}, + giveni={A\bibinitperiod}}}% + {{hash=b417efc17c6c4f027b621a1f48d8342e}{% + family={Li}, + familyi={L\bibinitperiod}, + given={Ruilong}, + giveni={R\bibinitperiod}}}% + {{hash=9a18985a5c8b8be23d0ec9de0a3c87b9}{% + family={Tancik}, + familyi={T\bibinitperiod}, + given={Matthew}, + giveni={M\bibinitperiod}}}% + {{hash=2620b9afd37cca5b9b7354f67c036d4d}{% + family={Li}, + familyi={L\bibinitperiod}, + given={Hao}, + giveni={H\bibinitperiod}}}% + {{hash=96e30bdc2cf905aa1109c46617b16252}{% + family={Ng}, + familyi={N\bibinitperiod}, + given={Ren}, + giveni={R\bibinitperiod}}}% + {{hash=6739ae1219e43710a65f37e9a5b5e902}{% + family={Kanazawa}, + familyi={K\bibinitperiod}, + given={Angjoo}, + giveni={A\bibinitperiod}}}% + } + \strng{namehash}{8a303a6a58c9fee7f11e0c90cc0ebcd6} + \strng{fullhash}{9efe82c0400b745a2e4e1b3e76382ea6} + \strng{bibnamehash}{9efe82c0400b745a2e4e1b3e76382ea6} + \strng{authorbibnamehash}{9efe82c0400b745a2e4e1b3e76382ea6} + \strng{authornamehash}{8a303a6a58c9fee7f11e0c90cc0ebcd6} + \strng{authorfullhash}{9efe82c0400b745a2e4e1b3e76382ea6} + \field{sortinit}{2} + \field{sortinithash}{8b555b3791beccb63322c22f3320aa9a} + \field{labelnamesource}{author} + \field{labeltitlesource}{title} + \field{booktitle}{Proceedings of the {{IEEE}}/{{CVF International Conference}} on {{Computer Vision}}} + \field{title}{Plenoctrees for Real-Time Rendering of Neural Radiance Fields} + \field{urlday}{13} + \field{urlmonth}{9} + \field{urlyear}{2024} + \field{year}{2021} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{5752\bibrangedash 5761} + \range{pages}{10} + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\relate_works\Yu et al_2021_Plenoctrees for real-time rendering of neural radiance fields.pdf + \endverb + \verb{urlraw} + \verb http://openaccess.thecvf.com/content/ICCV2021/html/Yu_PlenOctrees_for_Real-Time_Rendering_of_Neural_Radiance_Fields_ICCV_2021_paper.html + \endverb + \verb{url} + \verb http://openaccess.thecvf.com/content/ICCV2021/html/Yu_PlenOctrees_for_Real-Time_Rendering_of_Neural_Radiance_Fields_ICCV_2021_paper.html + \endverb + \keyw{⛔ No DOI found} + \endentry + \entry{fridovich-keilPlenoxelsRadianceFields2022}{inproceedings}{} + \name{author}{6}{}{% + {{hash=c631c1dd8a471790086b4fb6e8af17f2}{% + family={Fridovich-Keil}, + familyi={F\bibinithyphendelim K\bibinitperiod}, + given={Sara}, + giveni={S\bibinitperiod}}}% + {{hash=ff32d51d3d43688459e059d03b115eca}{% + family={Yu}, + familyi={Y\bibinitperiod}, + given={Alex}, + giveni={A\bibinitperiod}}}% + {{hash=9a18985a5c8b8be23d0ec9de0a3c87b9}{% + family={Tancik}, + familyi={T\bibinitperiod}, + given={Matthew}, + giveni={M\bibinitperiod}}}% + {{hash=87bb5d69ecd43c83ff6c1f229a3f059e}{% + family={Chen}, + familyi={C\bibinitperiod}, + given={Qinhong}, + giveni={Q\bibinitperiod}}}% + {{hash=3503059e1c0c778913607c87d8c5173a}{% + family={Recht}, + familyi={R\bibinitperiod}, + given={Benjamin}, + giveni={B\bibinitperiod}}}% + {{hash=6739ae1219e43710a65f37e9a5b5e902}{% + family={Kanazawa}, + familyi={K\bibinitperiod}, + given={Angjoo}, + giveni={A\bibinitperiod}}}% + } + \strng{namehash}{135f9445b6158d10e48a066eb09e1eb2} + \strng{fullhash}{13a051ca054a6ebd0c9c96b602778f7f} + \strng{bibnamehash}{13a051ca054a6ebd0c9c96b602778f7f} + \strng{authorbibnamehash}{13a051ca054a6ebd0c9c96b602778f7f} + \strng{authornamehash}{135f9445b6158d10e48a066eb09e1eb2} + \strng{authorfullhash}{13a051ca054a6ebd0c9c96b602778f7f} + \field{sortinit}{2} + \field{sortinithash}{8b555b3791beccb63322c22f3320aa9a} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{booktitle}{Proceedings of the {{IEEE}}/{{CVF}} Conference on Computer Vision and Pattern Recognition} + \field{shorttitle}{Plenoxels} + \field{title}{Plenoxels: {{Radiance}} Fields without Neural Networks} + \field{urlday}{3} + \field{urlmonth}{9} + \field{urlyear}{2024} + \field{year}{2022} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{5501\bibrangedash 5510} + \range{pages}{10} + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\relate_works\Fridovich-Keil et al_2022_Plenoxels.pdf + \endverb + \verb{urlraw} + \verb http://openaccess.thecvf.com/content/CVPR2022/html/Fridovich-Keil_Plenoxels_Radiance_Fields_Without_Neural_Networks_CVPR_2022_paper.html + \endverb + \verb{url} + \verb http://openaccess.thecvf.com/content/CVPR2022/html/Fridovich-Keil_Plenoxels_Radiance_Fields_Without_Neural_Networks_CVPR_2022_paper.html + \endverb + \keyw{⛔ No DOI found} + \endentry + \entry{keethaSplaTAMSplatTrack2024}{online}{} + \name{author}{7}{}{% + {{hash=1aee7c01f6f49b87068e78702c98ffcb}{% + family={Keetha}, + familyi={K\bibinitperiod}, + given={Nikhil}, + giveni={N\bibinitperiod}}}% + {{hash=c66eaaf176c294b1436c43321bab695a}{% + family={Karhade}, + familyi={K\bibinitperiod}, + given={Jay}, + giveni={J\bibinitperiod}}}% + {{hash=6ce2ec3bb10a1f4268e04ea143952547}{% + family={Jatavallabhula}, + familyi={J\bibinitperiod}, + given={Krishna\bibnamedelima Murthy}, + giveni={K\bibinitperiod\bibinitdelim M\bibinitperiod}}}% + {{hash=f3f4fb1e873ad6ec4b53d70078cfdb3a}{% + family={Yang}, + familyi={Y\bibinitperiod}, + given={Gengshan}, + giveni={G\bibinitperiod}}}% + {{hash=ad26bf16eeaee427e8e36bea35587822}{% + family={Scherer}, + familyi={S\bibinitperiod}, + given={Sebastian}, + giveni={S\bibinitperiod}}}% + {{hash=aedcd3831845183e070aef16857d83ea}{% + family={Ramanan}, + familyi={R\bibinitperiod}, + given={Deva}, + giveni={D\bibinitperiod}}}% + {{hash=8c336c193892d2f1d5a6c952eb5b1d4c}{% + family={Luiten}, + familyi={L\bibinitperiod}, + given={Jonathon}, + giveni={J\bibinitperiod}}}% + } + \strng{namehash}{327634b377ad49e4467ed06bfb213a85} + \strng{fullhash}{b2d05f0b8eeb825e0a63d870889ffdfd} + \strng{bibnamehash}{327634b377ad49e4467ed06bfb213a85} + \strng{authorbibnamehash}{327634b377ad49e4467ed06bfb213a85} + \strng{authornamehash}{327634b377ad49e4467ed06bfb213a85} + \strng{authorfullhash}{b2d05f0b8eeb825e0a63d870889ffdfd} + \field{sortinit}{2} + \field{sortinithash}{8b555b3791beccb63322c22f3320aa9a} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{abstract}{Dense simultaneous localization and mapping (SLAM) is crucial for robotics and augmented reality applications. However, current methods are often hampered by the non-volumetric or implicit way they represent a scene. This work introduces SplaTAM, an approach that, for the first time, leverages explicit volumetric representations, i.e., 3D Gaussians, to enable high-fidelity reconstruction from a single unposed RGB-D camera, surpassing the capabilities of existing methods. SplaTAM employs a simple online tracking and mapping system tailored to the underlying Gaussian representation. It utilizes a silhouette mask to elegantly capture the presence of scene density. This combination enables several benefits over prior representations, including fast rendering and dense optimization, quickly determining if areas have been previously mapped, and structured map expansion by adding more Gaussians. Extensive experiments show that SplaTAM achieves up to 2x superior performance in camera pose estimation, map construction, and novel-view synthesis over existing methods, paving the way for more immersive high-fidelity SLAM applications.} + \field{day}{16} + \field{eprintclass}{cs} + \field{eprinttype}{arXiv} + \field{month}{4} + \field{pubstate}{prepublished} + \field{shorttitle}{{{SplaTAM}}} + \field{title}{{{SplaTAM}}: {{Splat}}, {{Track}} \& {{Map 3D Gaussians}} for {{Dense RGB-D SLAM}}} + \field{urlday}{9} + \field{urlmonth}{6} + \field{urlyear}{2024} + \field{year}{2024} + \field{dateera}{ce} + \field{urldateera}{ce} + \verb{eprint} + \verb 2312.02126 + \endverb + \verb{file} + \verb C\:\\Users\\18317\\Zotero\\storage\\WUQKHF5Y\\Keetha et al. - 2024 - SplaTAM Splat, Track & Map 3D Gaussians for Dense.pdf;C\:\\Users\\18317\\Zotero\\storage\\5A6XMJDM\\2312.html + \endverb + \verb{urlraw} + \verb http://arxiv.org/abs/2312.02126 + \endverb + \verb{url} + \verb http://arxiv.org/abs/2312.02126 + \endverb + \keyw{Computer Science - Artificial Intelligence,Computer Science - Computer Vision and Pattern Recognition,Computer Science - Robotics} + \endentry + \entry{huCGSLAMEfficientDense2024}{online}{} + \name{author}{8}{}{% + {{hash=a646796e5765f8f44d0bd05cd156040b}{% + family={Hu}, + familyi={H\bibinitperiod}, + given={Jiarui}, + giveni={J\bibinitperiod}}}% + {{hash=8f66cf05cc1f53e0813865a26da33f60}{% + family={Chen}, + familyi={C\bibinitperiod}, + given={Xianhao}, + giveni={X\bibinitperiod}}}% + {{hash=6451fedde9c40387124a637057184390}{% + family={Feng}, + familyi={F\bibinitperiod}, + given={Boyin}, + giveni={B\bibinitperiod}}}% + {{hash=bb1182aa336dd5b4dba4a678d1bcc0ee}{% + family={Li}, + familyi={L\bibinitperiod}, + given={Guanglin}, + giveni={G\bibinitperiod}}}% + {{hash=482d5067f6f329ecac9a9c7f197280f2}{% + family={Yang}, + familyi={Y\bibinitperiod}, + given={Liangjing}, + giveni={L\bibinitperiod}}}% + {{hash=c384a68d4495c23c2753e5bbbad04be0}{% + family={Bao}, + familyi={B\bibinitperiod}, + given={Hujun}, + giveni={H\bibinitperiod}}}% + {{hash=0150ab01be364b2d61fa4e2f9e01e23b}{% + family={Zhang}, + familyi={Z\bibinitperiod}, + given={Guofeng}, + giveni={G\bibinitperiod}}}% + {{hash=152d039c26f4cbd157947af9ab63a949}{% + family={Cui}, + familyi={C\bibinitperiod}, + given={Zhaopeng}, + giveni={Z\bibinitperiod}}}% + } + \strng{namehash}{bc2e62901ad3a1e2d3f817e8bb7b2ce7} + \strng{fullhash}{f6ef02ac874f6fcfca2b681a72f58d3f} + \strng{bibnamehash}{bc2e62901ad3a1e2d3f817e8bb7b2ce7} + \strng{authorbibnamehash}{bc2e62901ad3a1e2d3f817e8bb7b2ce7} + \strng{authornamehash}{bc2e62901ad3a1e2d3f817e8bb7b2ce7} + \strng{authorfullhash}{f6ef02ac874f6fcfca2b681a72f58d3f} + \field{sortinit}{2} + \field{sortinithash}{8b555b3791beccb63322c22f3320aa9a} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{abstract}{Recently neural radiance fields (NeRF) have been widely exploited as 3D representations for dense simultaneous localization and mapping (SLAM). Despite their notable successes in surface modeling and novel view synthesis, existing NeRF-based methods are hindered by their computationally intensive and time-consuming volume rendering pipeline. This paper presents an efficient dense RGB-D SLAM system, i.e., CG-SLAM, based on a novel uncertainty-aware 3D Gaussian field with high consistency and geometric stability. Through an in-depth analysis of Gaussian Splatting, we propose several techniques to construct a consistent and stable 3D Gaussian field suitable for tracking and mapping. Additionally, a novel depth uncertainty model is proposed to ensure the selection of valuable Gaussian primitives during optimization, thereby improving tracking efficiency and accuracy. Experiments on various datasets demonstrate that CG-SLAM achieves superior tracking and mapping performance with a notable tracking speed of up to 15 Hz. We will make our source code publicly available. Project page: https://zju3dv.github.io/cg-slam.} + \field{day}{24} + \field{eprintclass}{cs} + \field{eprinttype}{arXiv} + \field{month}{3} + \field{pubstate}{prepublished} + \field{shorttitle}{{{CG-SLAM}}} + \field{title}{{{CG-SLAM}}: {{Efficient Dense RGB-D SLAM}} in a {{Consistent Uncertainty-aware 3D Gaussian Field}}} + \field{urlday}{16} + \field{urlmonth}{7} + \field{urlyear}{2024} + \field{year}{2024} + \field{dateera}{ce} + \field{urldateera}{ce} + \verb{eprint} + \verb 2403.16095 + \endverb + \verb{file} + \verb C\:\\Users\\18317\\DevSpace\\DocHub\\essay\\zotero\\attachments\\gs_spalt\\Hu et al_2024_CG-SLAM3.pdf;C\:\\Users\\18317\\Zotero\\storage\\EKRNMV59\\2403.html + \endverb + \verb{urlraw} + \verb http://arxiv.org/abs/2403.16095 + \endverb + \verb{url} + \verb http://arxiv.org/abs/2403.16095 + \endverb + \keyw{Computer Science - Computer Vision and Pattern Recognition,Computer Science - Robotics} + \endentry + \entry{beslMethodRegistration3shapes1992}{inproceedings}{} + \name{author}{2}{}{% + {{hash=c86e978974a8b442b51fbaa04acd2e0b}{% + family={Besl}, + familyi={B\bibinitperiod}, + given={Paul\bibnamedelima J.}, + giveni={P\bibinitperiod\bibinitdelim J\bibinitperiod}}}% + {{hash=069445b230220d610c916f4272b33ced}{% + family={McKay}, + familyi={M\bibinitperiod}, + given={Neil\bibnamedelima D.}, + giveni={N\bibinitperiod\bibinitdelim D\bibinitperiod}}}% + } + \list{publisher}{1}{% + {Spie}% + } + \strng{namehash}{46f98bbe16698f10efa6e1b762574b99} + \strng{fullhash}{46f98bbe16698f10efa6e1b762574b99} + \strng{bibnamehash}{46f98bbe16698f10efa6e1b762574b99} + \strng{authorbibnamehash}{46f98bbe16698f10efa6e1b762574b99} + \strng{authornamehash}{46f98bbe16698f10efa6e1b762574b99} + \strng{authorfullhash}{46f98bbe16698f10efa6e1b762574b99} + \field{sortinit}{3} + \field{sortinithash}{ad6fe7482ffbd7b9f99c9e8b5dccd3d7} + \field{labelnamesource}{author} + \field{labeltitlesource}{title} + \field{booktitle}{Sensor Fusion {{IV}}: Control Paradigms and Data Structures} + \field{title}{Method for Registration of 3-{{D}} Shapes} + \field{urlday}{13} + \field{urlmonth}{9} + \field{urlyear}{2024} + \field{volume}{1611} + \field{year}{1992} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{586\bibrangedash 606} + \range{pages}{21} + \verb{doi} + \verb 10.1117/12.57955 + \endverb + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\relate_works\Besl_McKay_1992_Method for registration of 3-D shapes.pdf + \endverb + \verb{urlraw} + \verb https://www.spiedigitallibrary.org/conference-proceedings-of-spie/1611/1/Method-for-registration-of-3-D-shapes/10.1117/12.57955.short + \endverb + \verb{url} + \verb https://www.spiedigitallibrary.org/conference-proceedings-of-spie/1611/1/Method-for-registration-of-3-D-shapes/10.1117/12.57955.short + \endverb + \endentry + \entry{segalGeneralizedicp2009a}{inproceedings}{} + \name{author}{3}{}{% + {{hash=f301e710afb9806e47054d6bb1be796f}{% + family={Segal}, + familyi={S\bibinitperiod}, + given={Aleksandr}, + giveni={A\bibinitperiod}}}% + {{hash=b87dffa311fa694fb509a2535cbdc408}{% + family={Haehnel}, + familyi={H\bibinitperiod}, + given={Dirk}, + giveni={D\bibinitperiod}}}% + {{hash=5112ea130eadef0764fce0145b8b9f91}{% + family={Thrun}, + familyi={T\bibinitperiod}, + given={Sebastian}, + giveni={S\bibinitperiod}}}% + } + \list{publisher}{1}{% + {Seattle, WA}% + } + \strng{namehash}{49bdf2e1ff1c0b17f77e99f8f17a4d8b} + \strng{fullhash}{1240901b76de1ada6eca66daf26d5dd1} + \strng{bibnamehash}{1240901b76de1ada6eca66daf26d5dd1} + \strng{authorbibnamehash}{1240901b76de1ada6eca66daf26d5dd1} + \strng{authornamehash}{49bdf2e1ff1c0b17f77e99f8f17a4d8b} + \strng{authorfullhash}{1240901b76de1ada6eca66daf26d5dd1} + \field{sortinit}{3} + \field{sortinithash}{ad6fe7482ffbd7b9f99c9e8b5dccd3d7} + \field{labelnamesource}{author} + \field{labeltitlesource}{title} + \field{booktitle}{Robotics: Science and Systems} + \field{number}{4} + \field{title}{Generalized-Icp.} + \field{urlday}{20} + \field{urlmonth}{5} + \field{urlyear}{2024} + \field{volume}{2} + \field{year}{2009} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{435} + \range{pages}{1} + \verb{doi} + \verb 10.15607/RSS.2009.V.021 + \endverb + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\EMPTY_COLLECTION_NAMESegal et al\Segal et al_2009_Generalized-icp.pdf + \endverb + \verb{urlraw} + \verb https://direct.mit.edu/books/edited-volume/chapter-pdf/2277340/9780262289801_cau.pdf + \endverb + \verb{url} + \verb https://direct.mit.edu/books/edited-volume/chapter-pdf/2277340/9780262289801_cau.pdf + \endverb + \endentry + \entry{parkColoredPointCloud2017}{inproceedings}{} + \name{author}{3}{}{% + {{hash=4601943d47b90174b1d991712b65af21}{% + family={Park}, + familyi={P\bibinitperiod}, + given={Jaesik}, + giveni={J\bibinitperiod}}}% + {{hash=fc2bd85da63b3d182bb3b3433b80a17f}{% + family={Zhou}, + familyi={Z\bibinitperiod}, + given={Qian-Yi}, + giveni={Q\bibinithyphendelim Y\bibinitperiod}}}% + {{hash=9025300f43a193cfed8e10e924f2a117}{% + family={Koltun}, + familyi={K\bibinitperiod}, + given={Vladlen}, + giveni={V\bibinitperiod}}}% + } + \strng{namehash}{066b7539fb4ab7df7544bbd723faa825} + \strng{fullhash}{551c608f0b29a1ba277a091bf4755bbe} + \strng{bibnamehash}{551c608f0b29a1ba277a091bf4755bbe} + \strng{authorbibnamehash}{551c608f0b29a1ba277a091bf4755bbe} + \strng{authornamehash}{066b7539fb4ab7df7544bbd723faa825} + \strng{authorfullhash}{551c608f0b29a1ba277a091bf4755bbe} + \field{sortinit}{3} + \field{sortinithash}{ad6fe7482ffbd7b9f99c9e8b5dccd3d7} + \field{labelnamesource}{author} + \field{labeltitlesource}{title} + \field{booktitle}{Proceedings of the {{IEEE}} International Conference on Computer Vision} + \field{title}{Colored Point Cloud Registration Revisited} + \field{urlday}{23} + \field{urlmonth}{5} + \field{urlyear}{2024} + \field{year}{2017} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{143\bibrangedash 152} + \range{pages}{10} + \verb{file} + \verb C:\Users\18317\Zotero\storage\NIWZJR2Y\Park et al_2017_Colored point cloud registration revisited-dual-translated.pdf + \endverb + \verb{urlraw} + \verb http://openaccess.thecvf.com/content_iccv_2017/html/Park_Colored_Point_Cloud_ICCV_2017_paper.html + \endverb + \verb{url} + \verb http://openaccess.thecvf.com/content_iccv_2017/html/Park_Colored_Point_Cloud_ICCV_2017_paper.html + \endverb + \keyw{💫} + \endentry + \entry{steinbruckerRealtimeVisualOdometry2011}{inproceedings}{} + \name{author}{3}{}{% + {{hash=c9ab5a637c8b26cc9a3011c81f1cddd2}{% + family={Steinbrücker}, + familyi={S\bibinitperiod}, + given={Frank}, + giveni={F\bibinitperiod}}}% + {{hash=9e291584a3b29b27c88ecfe0a566274c}{% + family={Sturm}, + familyi={S\bibinitperiod}, + given={Jürgen}, + giveni={J\bibinitperiod}}}% + {{hash=1bd2b6b6ca2fc15a90f164070b626131}{% + family={Cremers}, + familyi={C\bibinitperiod}, + given={Daniel}, + giveni={D\bibinitperiod}}}% + } + \list{publisher}{1}{% + {IEEE}% + } + \strng{namehash}{7cc5b4309280651ef45cdfc5c5c68bad} + \strng{fullhash}{f640ca48ba37fc08ce9706957846b982} + \strng{bibnamehash}{f640ca48ba37fc08ce9706957846b982} + \strng{authorbibnamehash}{f640ca48ba37fc08ce9706957846b982} + \strng{authornamehash}{7cc5b4309280651ef45cdfc5c5c68bad} + \strng{authorfullhash}{f640ca48ba37fc08ce9706957846b982} + \field{sortinit}{3} + \field{sortinithash}{ad6fe7482ffbd7b9f99c9e8b5dccd3d7} + \field{labelnamesource}{author} + \field{labeltitlesource}{title} + \field{booktitle}{2011 {{IEEE}} International Conference on Computer Vision Workshops ({{ICCV Workshops}})} + \field{title}{Real-Time Visual Odometry from Dense {{RGB-D}} Images} + \field{urlday}{20} + \field{urlmonth}{8} + \field{urlyear}{2024} + \field{year}{2011} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{719\bibrangedash 722} + \range{pages}{4} + \verb{doi} + \verb 10.1109/ICCVW.2011.6130321 + \endverb + \verb{urlraw} + \verb https://ieeexplore.ieee.org/abstract/document/6130321/ + \endverb + \verb{url} + \verb https://ieeexplore.ieee.org/abstract/document/6130321/ + \endverb + \keyw{nosource} + \endentry + \entry{pomerleauComparingICPVariants2013}{article}{} + \name{author}{4}{}{% + {{hash=f220f9566be4f464ab62f74f22203b54}{% + family={Pomerleau}, + familyi={P\bibinitperiod}, + given={François}, + giveni={F\bibinitperiod}}}% + {{hash=05c5b6f76d792212ca5d644bb169272d}{% + family={Colas}, + familyi={C\bibinitperiod}, + given={Francis}, + giveni={F\bibinitperiod}}}% + {{hash=6ee8958bbf32527489012d8ca7c95ee3}{% + family={Siegwart}, + familyi={S\bibinitperiod}, + given={Roland}, + giveni={R\bibinitperiod}}}% + {{hash=6746a55bf5327ab107e8086d745e74f5}{% + family={Magnenat}, + familyi={M\bibinitperiod}, + given={Stéphane}, + giveni={S\bibinitperiod}}}% + } + \strng{namehash}{1e56ad13dbd841ad5970bb119dce8549} + \strng{fullhash}{99e70ee7c80bfbe9778fdf343f6ecd22} + \strng{bibnamehash}{99e70ee7c80bfbe9778fdf343f6ecd22} + \strng{authorbibnamehash}{99e70ee7c80bfbe9778fdf343f6ecd22} + \strng{authornamehash}{1e56ad13dbd841ad5970bb119dce8549} + \strng{authorfullhash}{99e70ee7c80bfbe9778fdf343f6ecd22} + \field{sortinit}{3} + \field{sortinithash}{ad6fe7482ffbd7b9f99c9e8b5dccd3d7} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{issn}{0929-5593, 1573-7527} + \field{journaltitle}{Autonomous Robots} + \field{langid}{english} + \field{month}{4} + \field{number}{3} + \field{shortjournal}{Auton Robot} + \field{shorttitle}{Comparing {{ICP}} Variants on Real-World Data Sets} + \field{title}{Comparing {{ICP}} Variants on Real-World Data Sets: {{Open-source}} Library and Experimental Protocol} + \field{urlday}{26} + \field{urlmonth}{9} + \field{urlyear}{2024} + \field{volume}{34} + \field{year}{2013} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{133\bibrangedash 148} + \range{pages}{16} + \verb{doi} + \verb 10.1007/s10514-013-9327-2 + \endverb + \verb{file} + \verb C\:\\Users\\18317\\DevSpace\\DocHub\\essay\\zotero\\attachments\\gs_spalt\\relate_works\\Pomerleau et al_2013_Comparing ICP variants on real-world data sets.pdf;C\:\\Users\\18317\\DevSpace\\DocHub\\essay\\zotero\\attachments\\gs_spalt\\relate_works\\Pomerleau et al_2013_Comparing ICP variants on real-world data sets2.pdf + \endverb + \verb{urlraw} + \verb http://link.springer.com/10.1007/s10514-013-9327-2 + \endverb + \verb{url} + \verb http://link.springer.com/10.1007/s10514-013-9327-2 + \endverb + \endentry + \entry{kuipersQuaternionsRotationSequences1999}{book}{} + \name{author}{1}{}{% + {{hash=6f6cb9d14f282595edbde9b8a9b025e2}{% + family={Kuipers}, + familyi={K\bibinitperiod}, + given={Jack\bibnamedelima B.}, + giveni={J\bibinitperiod\bibinitdelim B\bibinitperiod}}}% + } + \list{publisher}{1}{% + {Princeton university press}% + } + \strng{namehash}{6f6cb9d14f282595edbde9b8a9b025e2} + \strng{fullhash}{6f6cb9d14f282595edbde9b8a9b025e2} + \strng{bibnamehash}{6f6cb9d14f282595edbde9b8a9b025e2} + \strng{authorbibnamehash}{6f6cb9d14f282595edbde9b8a9b025e2} + \strng{authornamehash}{6f6cb9d14f282595edbde9b8a9b025e2} + \strng{authorfullhash}{6f6cb9d14f282595edbde9b8a9b025e2} + \field{sortinit}{3} + \field{sortinithash}{ad6fe7482ffbd7b9f99c9e8b5dccd3d7} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{shorttitle}{Quaternions and Rotation Sequences} + \field{title}{Quaternions and Rotation Sequences: A Primer with Applications to Orbits, Aerospace, and Virtual Reality} + \field{urlday}{13} + \field{urlmonth}{9} + \field{urlyear}{2024} + \field{year}{1999} + \field{dateera}{ce} + \field{urldateera}{ce} + \verb{file} + \verb C\:\\Users\\18317\\DevSpace\\DocHub\\essay\\zotero\\attachments\\gs_spalt\\relate_works\\Kuipers_1999_Quaternions and rotation sequences.pdf;C\:\\Users\\18317\\Zotero\\storage\\6MZ26W74\\Kuipers - 1999 - Quaternions and rotation sequences a primer with .pdf + \endverb + \verb{urlraw} + \verb https://books.google.com/books?hl=en&lr=&id=_Og9DwAAQBAJ&oi=fnd&pg=PR21&dq=kuipers+Quaternions+Rotation+Sequences+1999&ots=t3I3ky6Zut&sig=o3xQEsvNHM-t2AIySNcaxBhSc_I + \endverb + \verb{url} + \verb https://books.google.com/books?hl=en&lr=&id=_Og9DwAAQBAJ&oi=fnd&pg=PR21&dq=kuipers+Quaternions+Rotation+Sequences+1999&ots=t3I3ky6Zut&sig=o3xQEsvNHM-t2AIySNcaxBhSc_I + \endverb + \endentry + \entry{zwickerEWASplatting2002}{article}{} + \name{author}{4}{}{% + {{hash=f8b89b955dad449fd9254a354c6041d2}{% + family={Zwicker}, + familyi={Z\bibinitperiod}, + given={Matthias}, + giveni={M\bibinitperiod}}}% + {{hash=b996af7e283eefe990aabf24ef52e032}{% + family={Pfister}, + familyi={P\bibinitperiod}, + given={Hanspeter}, + giveni={H\bibinitperiod}}}% + {{hash=a86a443dd3dac8a6e34fa2ab939d2779}{% + family={Van\bibnamedelima Baar}, + familyi={V\bibinitperiod\bibinitdelim B\bibinitperiod}, + given={Jeroen}, + giveni={J\bibinitperiod}}}% + {{hash=89835485884945e38712de8162277435}{% + family={Gross}, + familyi={G\bibinitperiod}, + given={Markus}, + giveni={M\bibinitperiod}}}% + } + \list{publisher}{1}{% + {IEEE}% + } + \strng{namehash}{2e41ba32ecf73baf435ec7c653bc332b} + \strng{fullhash}{c4719443b6f94afe9feb78854d13a13f} + \strng{bibnamehash}{c4719443b6f94afe9feb78854d13a13f} + \strng{authorbibnamehash}{c4719443b6f94afe9feb78854d13a13f} + \strng{authornamehash}{2e41ba32ecf73baf435ec7c653bc332b} + \strng{authorfullhash}{c4719443b6f94afe9feb78854d13a13f} + \field{sortinit}{4} + \field{sortinithash}{9381316451d1b9788675a07e972a12a7} + \field{labelnamesource}{author} + \field{labeltitlesource}{title} + \field{journaltitle}{IEEE Transactions on Visualization and Computer Graphics} + \field{number}{3} + \field{title}{{{EWA}} Splatting} + \field{urlday}{20} + \field{urlmonth}{7} + \field{urlyear}{2024} + \field{volume}{8} + \field{year}{2002} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{223\bibrangedash 238} + \range{pages}{16} + \verb{doi} + \verb 10.1109/TVCG.2002.1021576 + \endverb + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\Zwicker et al_2002_EWA splatting.pdf + \endverb + \verb{urlraw} + \verb https://ieeexplore.ieee.org/abstract/document/1021576/ + \endverb + \verb{url} + \verb https://ieeexplore.ieee.org/abstract/document/1021576/ + \endverb + \endentry + \entry{kanopoulosDesignImageEdge1988}{article}{} + \name{author}{3}{}{% + {{hash=85a6cf1e3dec085321f0b999ae485311}{% + family={Kanopoulos}, + familyi={K\bibinitperiod}, + given={Nick}, + giveni={N\bibinitperiod}}}% + {{hash=06c924091fbb57e33b167243fcb9b45c}{% + family={Vasanthavada}, + familyi={V\bibinitperiod}, + given={Nagesh}, + giveni={N\bibinitperiod}}}% + {{hash=232f7a7dc0e96f563e4033bc80e64c85}{% + family={Baker}, + familyi={B\bibinitperiod}, + given={Robert\bibnamedelima L.}, + giveni={R\bibinitperiod\bibinitdelim L\bibinitperiod}}}% + } + \list{publisher}{1}{% + {IEEE}% + } + \strng{namehash}{ae36959f41470f9b537cadb9ff5d0006} + \strng{fullhash}{9d7b455e4ba7eaf22a062b3fc9edd8df} + \strng{bibnamehash}{9d7b455e4ba7eaf22a062b3fc9edd8df} + \strng{authorbibnamehash}{9d7b455e4ba7eaf22a062b3fc9edd8df} + \strng{authornamehash}{ae36959f41470f9b537cadb9ff5d0006} + \strng{authorfullhash}{9d7b455e4ba7eaf22a062b3fc9edd8df} + \field{sortinit}{4} + \field{sortinithash}{9381316451d1b9788675a07e972a12a7} + \field{labelnamesource}{author} + \field{labeltitlesource}{title} + \field{journaltitle}{IEEE Journal of solid-state circuits} + \field{number}{2} + \field{title}{Design of an Image Edge Detection Filter Using the {{Sobel}} Operator} + \field{urlday}{9} + \field{urlmonth}{8} + \field{urlyear}{2024} + \field{volume}{23} + \field{year}{1988} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{358\bibrangedash 367} + \range{pages}{10} + \verb{doi} + \verb 10.1109/4.996 + \endverb + \verb{urlraw} + \verb https://ieeexplore.ieee.org/abstract/document/996/ + \endverb + \verb{url} + \verb https://ieeexplore.ieee.org/abstract/document/996/ + \endverb + \keyw{nosource} + \endentry + \entry{kingmaAdamMethodStochastic2014}{unpublished}{} + \name{author}{1}{}{% + {{hash=b6fbd171848aad4edf3925543f1f1522}{% + family={Kingma}, + familyi={K\bibinitperiod}, + given={Diederik\bibnamedelima P.}, + giveni={D\bibinitperiod\bibinitdelim P\bibinitperiod}}}% + } + \strng{namehash}{b6fbd171848aad4edf3925543f1f1522} + \strng{fullhash}{b6fbd171848aad4edf3925543f1f1522} + \strng{bibnamehash}{b6fbd171848aad4edf3925543f1f1522} + \strng{authorbibnamehash}{b6fbd171848aad4edf3925543f1f1522} + \strng{authornamehash}{b6fbd171848aad4edf3925543f1f1522} + \strng{authorfullhash}{b6fbd171848aad4edf3925543f1f1522} + \field{sortinit}{4} + \field{sortinithash}{9381316451d1b9788675a07e972a12a7} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{eprinttype}{arXiv} + \field{shorttitle}{Adam} + \field{title}{Adam: {{A}} Method for Stochastic Optimization} + \field{year}{2014} + \field{dateera}{ce} + \verb{eprint} + \verb 1412.6980 + \endverb + \keyw{⛔ No DOI found} + \endentry + \entry{straubReplicaDatasetDigital2019}{online}{} + \name{author}{30}{}{% + {{hash=449da8c93866a386da0e55cfd97629b1}{% + family={Straub}, + familyi={S\bibinitperiod}, + given={Julian}, + giveni={J\bibinitperiod}}}% + {{hash=547364b3c154f5f2f7d53a4542e8e7c4}{% + family={Whelan}, + familyi={W\bibinitperiod}, + given={Thomas}, + giveni={T\bibinitperiod}}}% + {{hash=85bb1803c3b8c449445e27a74d0465ac}{% + family={Ma}, + familyi={M\bibinitperiod}, + given={Lingni}, + giveni={L\bibinitperiod}}}% + {{hash=0733d91fbf2a30c44e626cd56dc54c4d}{% + family={Chen}, + familyi={C\bibinitperiod}, + given={Yufan}, + giveni={Y\bibinitperiod}}}% + {{hash=9e9515b35f1161d2ea10f097e0222638}{% + family={Wijmans}, + familyi={W\bibinitperiod}, + given={Erik}, + giveni={E\bibinitperiod}}}% + {{hash=43d4028764737f96da619eec9e803be5}{% + family={Green}, + familyi={G\bibinitperiod}, + given={Simon}, + giveni={S\bibinitperiod}}}% + {{hash=b018e029e1ebc3b7a1d68a990c14414c}{% + family={Engel}, + familyi={E\bibinitperiod}, + given={Jakob\bibnamedelima J.}, + giveni={J\bibinitperiod\bibinitdelim J\bibinitperiod}}}% + {{hash=3bcf9140fcfbbb3fa5cdedbcb3380ee8}{% + family={Mur-Artal}, + familyi={M\bibinithyphendelim A\bibinitperiod}, + given={Raul}, + giveni={R\bibinitperiod}}}% + {{hash=30b9ddd4c1532e2e70298c97d606f54a}{% + family={Ren}, + familyi={R\bibinitperiod}, + given={Carl}, + giveni={C\bibinitperiod}}}% + {{hash=7979cf64afeedc0d5cce42ec332196cc}{% + family={Verma}, + familyi={V\bibinitperiod}, + given={Shobhit}, + giveni={S\bibinitperiod}}}% + {{hash=2e2efc330f9042584ac249de0665ce0c}{% + family={Clarkson}, + familyi={C\bibinitperiod}, + given={Anton}, + giveni={A\bibinitperiod}}}% + {{hash=a76494925059ff5b3543a093d868f6b7}{% + family={Yan}, + familyi={Y\bibinitperiod}, + given={Mingfei}, + giveni={M\bibinitperiod}}}% + {{hash=801e533aca932dccbe47ef24ce6844e6}{% + family={Budge}, + familyi={B\bibinitperiod}, + given={Brian}, + giveni={B\bibinitperiod}}}% + {{hash=f822255ffbf7573ee86e3ea03b883f69}{% + family={Yan}, + familyi={Y\bibinitperiod}, + given={Yajie}, + giveni={Y\bibinitperiod}}}% + {{hash=145354182208a8b3bdfe4a31a90ca58a}{% + family={Pan}, + familyi={P\bibinitperiod}, + given={Xiaqing}, + giveni={X\bibinitperiod}}}% + {{hash=3fc509f7d75521c6ccf50cc6fb7916bc}{% + family={Yon}, + familyi={Y\bibinitperiod}, + given={June}, + giveni={J\bibinitperiod}}}% + {{hash=0851ff9e983c5d81b64f367f3d198e3f}{% + family={Zou}, + familyi={Z\bibinitperiod}, + given={Yuyang}, + giveni={Y\bibinitperiod}}}% + {{hash=0e721e23e818845dbcf63cec1af3a6d2}{% + family={Leon}, + familyi={L\bibinitperiod}, + given={Kimberly}, + giveni={K\bibinitperiod}}}% + {{hash=60933f16bcf58e4a6761006e8ab3d20a}{% + family={Carter}, + familyi={C\bibinitperiod}, + given={Nigel}, + giveni={N\bibinitperiod}}}% + {{hash=f313387efc828661f1f1eaec4132c279}{% + family={Briales}, + familyi={B\bibinitperiod}, + given={Jesus}, + giveni={J\bibinitperiod}}}% + {{hash=20afb38370996f7859c54d4ea00f69ef}{% + family={Gillingham}, + familyi={G\bibinitperiod}, + given={Tyler}, + giveni={T\bibinitperiod}}}% + {{hash=d4f0473d41e0f621d04087560afd4065}{% + family={Mueggler}, + familyi={M\bibinitperiod}, + given={Elias}, + giveni={E\bibinitperiod}}}% + {{hash=9b336ad198387bad7dacf615c036295a}{% + family={Pesqueira}, + familyi={P\bibinitperiod}, + given={Luis}, + giveni={L\bibinitperiod}}}% + {{hash=bad0a30ec9610ab126c95e2785c806c4}{% + family={Savva}, + familyi={S\bibinitperiod}, + given={Manolis}, + giveni={M\bibinitperiod}}}% + {{hash=d0a408119272ca1bd49b625e3b927d05}{% + family={Batra}, + familyi={B\bibinitperiod}, + given={Dhruv}, + giveni={D\bibinitperiod}}}% + {{hash=1ae3b92892b9bd4f6e3bd54021ba0fbe}{% + family={Strasdat}, + familyi={S\bibinitperiod}, + given={Hauke\bibnamedelima M.}, + giveni={H\bibinitperiod\bibinitdelim M\bibinitperiod}}}% + {{hash=90d5753b2165d531b4206ac54756b857}{% + family={De\bibnamedelima Nardi}, + familyi={D\bibinitperiod\bibinitdelim N\bibinitperiod}, + given={Renzo}, + giveni={R\bibinitperiod}}}% + {{hash=fe9c4886922d0dc4b62709fdad478c05}{% + family={Goesele}, + familyi={G\bibinitperiod}, + given={Michael}, + giveni={M\bibinitperiod}}}% + {{hash=e7c982cac058f95aeb5e1e584f764c82}{% + family={Lovegrove}, + familyi={L\bibinitperiod}, + given={Steven}, + giveni={S\bibinitperiod}}}% + {{hash=239670d06a0c3f1468b54a7264838ce6}{% + family={Newcombe}, + familyi={N\bibinitperiod}, + given={Richard}, + giveni={R\bibinitperiod}}}% + } + \strng{namehash}{23704b5d33a408d95e7a187d9810e87e} + \strng{fullhash}{634d54df49ae2207ad62722508cba4cd} + \strng{bibnamehash}{23704b5d33a408d95e7a187d9810e87e} + \strng{authorbibnamehash}{23704b5d33a408d95e7a187d9810e87e} + \strng{authornamehash}{23704b5d33a408d95e7a187d9810e87e} + \strng{authorfullhash}{634d54df49ae2207ad62722508cba4cd} + \field{sortinit}{4} + \field{sortinithash}{9381316451d1b9788675a07e972a12a7} + \field{labelnamesource}{author} + \field{labeltitlesource}{shorttitle} + \field{abstract}{We introduce Replica, a dataset of 18 highly photo-realistic 3D indoor scene reconstructions at room and building scale. Each scene consists of a dense mesh, high-resolution high-dynamic-range (HDR) textures, per-primitive semantic class and instance information, and planar mirror and glass reflectors. The goal of Replica is to enable machine learning (ML) research that relies on visually, geometrically, and semantically realistic generative models of the world - for instance, egocentric computer vision, semantic segmentation in 2D and 3D, geometric inference, and the development of embodied agents (virtual robots) performing navigation, instruction following, and question answering. Due to the high level of realism of the renderings from Replica, there is hope that ML systems trained on Replica may transfer directly to real world image and video data. Together with the data, we are releasing a minimal C++ SDK as a starting point for working with the Replica dataset. In addition, Replica is `Habitat-compatible', i.e. can be natively used with AI Habitat for training and testing embodied agents.} + \field{day}{13} + \field{eprintclass}{cs, eess} + \field{eprinttype}{arXiv} + \field{month}{6} + \field{pubstate}{prepublished} + \field{shorttitle}{The {{Replica Dataset}}} + \field{title}{The {{Replica Dataset}}: {{A Digital Replica}} of {{Indoor Spaces}}} + \field{urlday}{10} + \field{urlmonth}{8} + \field{urlyear}{2024} + \field{year}{2019} + \field{dateera}{ce} + \field{urldateera}{ce} + \verb{eprint} + \verb 1906.05797 + \endverb + \verb{file} + \verb C\:\\Users\\18317\\DevSpace\\DocHub\\essay\\zotero\\attachments\\gs_spalt\\experiments\\Straub et al_2019_The Replica Dataset.pdf;C\:\\Users\\18317\\Zotero\\storage\\VRTGI9PS\\1906.html + \endverb + \verb{urlraw} + \verb http://arxiv.org/abs/1906.05797 + \endverb + \verb{url} + \verb http://arxiv.org/abs/1906.05797 + \endverb + \keyw{Computer Science - Computer Vision and Pattern Recognition,Computer Science - Graphics,Electrical Engineering and Systems Science - Image and Video Processing} + \endentry + \entry{sturmBenchmarkEvaluationRGBD2012}{inproceedings}{} + \name{author}{5}{}{% + {{hash=9e291584a3b29b27c88ecfe0a566274c}{% + family={Sturm}, + familyi={S\bibinitperiod}, + given={Jürgen}, + giveni={J\bibinitperiod}}}% + {{hash=9d6110950a50d69179aefe99b3abc81a}{% + family={Engelhard}, + familyi={E\bibinitperiod}, + given={Nikolas}, + giveni={N\bibinitperiod}}}% + {{hash=4858aa954321596e8c1c81daac0271dd}{% + family={Endres}, + familyi={E\bibinitperiod}, + given={Felix}, + giveni={F\bibinitperiod}}}% + {{hash=98f4ab1bc2ac191d8f8a0651315ce9c0}{% + family={Burgard}, + familyi={B\bibinitperiod}, + given={Wolfram}, + giveni={W\bibinitperiod}}}% + {{hash=1bd2b6b6ca2fc15a90f164070b626131}{% + family={Cremers}, + familyi={C\bibinitperiod}, + given={Daniel}, + giveni={D\bibinitperiod}}}% + } + \list{publisher}{1}{% + {IEEE}% + } + \strng{namehash}{f2452d6a1e4ac6399e084b1a065ee7da} + \strng{fullhash}{fda811f69253ca27ebab9abbbf7b757f} + \strng{bibnamehash}{fda811f69253ca27ebab9abbbf7b757f} + \strng{authorbibnamehash}{fda811f69253ca27ebab9abbbf7b757f} + \strng{authornamehash}{f2452d6a1e4ac6399e084b1a065ee7da} + \strng{authorfullhash}{fda811f69253ca27ebab9abbbf7b757f} + \field{sortinit}{4} + \field{sortinithash}{9381316451d1b9788675a07e972a12a7} + \field{labelnamesource}{author} + \field{labeltitlesource}{title} + \field{booktitle}{2012 {{IEEE}}/{{RSJ}} International Conference on Intelligent Robots and Systems} + \field{title}{A Benchmark for the Evaluation of {{RGB-D SLAM}} Systems} + \field{urlday}{10} + \field{urlmonth}{8} + \field{urlyear}{2024} + \field{year}{2012} + \field{dateera}{ce} + \field{urldateera}{ce} + \field{pages}{573\bibrangedash 580} + \range{pages}{8} + \verb{doi} + \verb 10.1109/IROS.2012.6385773 + \endverb + \verb{file} + \verb C:\Users\18317\DevSpace\DocHub\essay\zotero\attachments\gs_spalt\experiments\Sturm et al_2012_A benchmark for the evaluation of RGB-D SLAM systems.pdf + \endverb + \verb{urlraw} + \verb https://ieeexplore.ieee.org/abstract/document/6385773/ + \endverb + \verb{url} + \verb https://ieeexplore.ieee.org/abstract/document/6385773/ + \endverb + \endentry + \enddatalist +\endrefsection +\endinput + diff --git a/docs/submission/main.bcf b/docs/submission/main.bcf new file mode 100644 index 0000000..5f07a2e --- /dev/null +++ b/docs/submission/main.bcf @@ -0,0 +1,2483 @@ + + + + + + output_encoding + utf8 + + + input_encoding + utf8 + + + debug + 0 + + + mincrossrefs + 2 + + + minxrefs + 2 + + + sortcase + 1 + + + sortupper + 1 + + + + + + + alphaothers + + + + + extradatecontext + labelname + labeltitle + + + labelalpha + 0 + + + labelnamespec + shortauthor + author + shorteditor + editor + translator + + + labeltitle + 0 + + + labeltitlespec + shorttitle + title + maintitle + + + labeltitleyear + 0 + + + labeldateparts + 0 + + + labeldatespec + date + year + eventdate + origdate + urldate + nodate + + + julian + 0 + + + gregorianstart + 1582-10-15 + + + maxalphanames + 3 + + + maxbibnames + 6 + + + maxcitenames + 2 + + + maxsortnames + 6 + + + maxitems + 3 + + + minalphanames + 1 + + + minbibnames + 1 + + + mincitenames + 1 + + + minsortnames + 1 + + + minitems + 1 + + + nohashothers + 0 + + + noroman + 0 + + + nosortothers + 0 + + + pluralothers + 0 + + + singletitle + 0 + + + skipbib + 0 + + + skipbiblist + 0 + + + skiplab + 0 + + + sortalphaothers + + + + + sortlocale + english + + + sortingtemplatename + none + + + sortsets + 0 + + + uniquelist + false + + + uniquename + false + + + uniqueprimaryauthor + 0 + + + uniquetitle + 0 + + + uniquebaretitle + 0 + + + uniquework + 0 + + + useprefix + 0 + + + useafterword + 1 + + + useannotator + 1 + + + useauthor + 1 + + + usebookauthor + 1 + + + usecommentator + 1 + + + useeditor + 1 + + + useeditora + 1 + + + useeditorb + 1 + + + useeditorc + 1 + + + useforeword + 1 + + + useholder + 1 + + + useintroduction + 1 + + + usenamea + 1 + + + usenameb + 1 + + + usenamec + 1 + + + usetranslator + 0 + + + useshortauthor + 1 + + + useshorteditor + 1 + + + + + + extradatecontext + labelname + labeltitle + + + labelalpha + 0 + + + labelnamespec + shortauthor + author + shorteditor + editor + translator + + + labeltitle + 0 + + + labeltitlespec + shorttitle + title + maintitle + + + labeltitleyear + 0 + + + labeldateparts + 0 + + + labeldatespec + date + year + eventdate + origdate + urldate + nodate + + + maxalphanames + 3 + + + maxbibnames + 6 + + + maxcitenames + 2 + + + maxsortnames + 6 + + + maxitems + 3 + + + minalphanames + 1 + + + minbibnames + 1 + + + mincitenames + 1 + + + minsortnames + 1 + + + minitems + 1 + + + nohashothers + 0 + + + noroman + 0 + + + nosortothers + 0 + + + singletitle + 0 + + + skipbib + 0 + + + skipbiblist + 0 + + + skiplab + 0 + + + uniquelist + false + + + uniquename + false + + + uniqueprimaryauthor + 0 + + + uniquetitle + 0 + + + uniquebaretitle + 0 + + + uniquework + 0 + + + useprefix + 0 + + + useafterword + 1 + + + useannotator + 1 + + + useauthor + 1 + + + usebookauthor + 1 + + + usecommentator + 1 + + + useeditor + 1 + + + useeditora + 1 + + + useeditorb + 1 + + + useeditorc + 1 + + + useforeword + 1 + + + useholder + 1 + + + useintroduction + 1 + + + usenamea + 1 + + + usenameb + 1 + + + usenamec + 1 + + + usetranslator + 0 + + + useshortauthor + 1 + + + useshorteditor + 1 + + + + + datamodel + labelalphanametemplate + labelalphatemplate + inheritance + translit + uniquenametemplate + sortingnamekeytemplate + sortingtemplate + extradatespec + extradatecontext + labelnamespec + labeltitlespec + labeldatespec + controlversion + alphaothers + sortalphaothers + presort + texencoding + bibencoding + sortingtemplatename + sortlocale + language + autolang + langhook + indexing + hyperref + backrefsetstyle + block + pagetracker + citecounter + citetracker + ibidtracker + idemtracker + opcittracker + loccittracker + labeldate + labeltime + dateera + date + time + eventdate + eventtime + origdate + origtime + urldate + urltime + alldatesusetime + alldates + alltimes + gregorianstart + autocite + notetype + uniquelist + uniquename + refsection + refsegment + citereset + sortlos + babel + datelabel + backrefstyle + arxiv + familyinits + giveninits + prefixinits + suffixinits + useafterword + useannotator + useauthor + usebookauthor + usecommentator + useeditor + useeditora + useeditorb + useeditorc + useforeword + useholder + useintroduction + usenamea + usenameb + usenamec + usetranslator + useshortauthor + useshorteditor + debug + loadfiles + safeinputenc + sortcase + sortupper + terseinits + abbreviate + dateabbrev + clearlang + sortcites + sortsets + backref + backreffloats + trackfloats + parentracker + labeldateusetime + datecirca + dateuncertain + dateusetime + eventdateusetime + origdateusetime + urldateusetime + julian + datezeros + timezeros + timezones + seconds + autopunct + punctfont + labelnumber + labelalpha + labeltitle + labeltitleyear + labeldateparts + pluralothers + nohashothers + nosortothers + noroman + singletitle + uniquetitle + uniquebaretitle + uniquework + uniqueprimaryauthor + defernumbers + locallabelwidth + bibwarn + useprefix + skipbib + skipbiblist + skiplab + dataonly + defernums + firstinits + sortfirstinits + sortgiveninits + labelyear + isbn + url + doi + eprint + related + subentry + dashed + bibtexcaseprotection + mincrossrefs + minxrefs + maxnames + minnames + maxbibnames + minbibnames + maxcitenames + mincitenames + maxsortnames + minsortnames + maxitems + minitems + maxalphanames + minalphanames + maxparens + dateeraauto + + + alphaothers + sortalphaothers + presort + indexing + citetracker + ibidtracker + idemtracker + opcittracker + loccittracker + uniquelist + uniquename + familyinits + giveninits + prefixinits + suffixinits + useafterword + useannotator + useauthor + usebookauthor + usecommentator + useeditor + useeditora + useeditorb + useeditorc + useforeword + useholder + useintroduction + usenamea + usenameb + usenamec + usetranslator + useshortauthor + useshorteditor + terseinits + abbreviate + dateabbrev + clearlang + labelnumber + labelalpha + labeltitle + labeltitleyear + labeldateparts + nohashothers + nosortothers + noroman + singletitle + uniquetitle + uniquebaretitle + uniquework + uniqueprimaryauthor + useprefix + skipbib + skipbiblist + skiplab + dataonly + skiplos + labelyear + isbn + url + doi + eprint + related + subentry + bibtexcaseprotection + labelalphatemplate + translit + sortexclusion + sortinclusion + extradatecontext + labelnamespec + labeltitlespec + labeldatespec + maxnames + minnames + maxbibnames + minbibnames + maxcitenames + mincitenames + maxsortnames + minsortnames + maxitems + minitems + maxalphanames + minalphanames + + + noinherit + nametemplates + labelalphanametemplatename + uniquenametemplatename + sortingnamekeytemplatename + presort + indexing + citetracker + ibidtracker + idemtracker + opcittracker + loccittracker + uniquelist + uniquename + familyinits + giveninits + prefixinits + suffixinits + useafterword + useannotator + useauthor + usebookauthor + usecommentator + useeditor + useeditora + useeditorb + useeditorc + useforeword + useholder + useintroduction + usenamea + usenameb + usenamec + usetranslator + useshortauthor + useshorteditor + terseinits + abbreviate + dateabbrev + clearlang + labelnumber + labelalpha + labeltitle + labeltitleyear + labeldateparts + nohashothers + nosortothers + noroman + singletitle + uniquetitle + uniquebaretitle + uniquework + uniqueprimaryauthor + useprefix + skipbib + skipbiblist + skiplab + dataonly + skiplos + isbn + url + doi + eprint + related + subentry + bibtexcaseprotection + maxnames + minnames + maxbibnames + minbibnames + maxcitenames + mincitenames + maxsortnames + minsortnames + maxitems + minitems + maxalphanames + minalphanames + + + nametemplates + labelalphanametemplatename + uniquenametemplatename + sortingnamekeytemplatename + uniquelist + uniquename + familyinits + giveninits + prefixinits + suffixinits + terseinits + nohashothers + nosortothers + useprefix + + + nametemplates + labelalphanametemplatename + uniquenametemplatename + sortingnamekeytemplatename + uniquename + familyinits + giveninits + prefixinits + suffixinits + terseinits + useprefix + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + prefix + family + + + + + shorthand + label + labelname + labelname + + + year + + + + + + labelyear + year + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + prefix + family + given + + + + + prefix + family + + + given + + + suffix + + + prefix + + + mm + + + + sf,sm,sn,pf,pm,pn,pp + family,given,prefix,suffix + boolean,integer,string,xml + default,transliteration,transcription,translation + + + article + artwork + audio + bibnote + book + bookinbook + booklet + collection + commentary + customa + customb + customc + customd + custome + customf + dataset + inbook + incollection + inproceedings + inreference + image + jurisdiction + legal + legislation + letter + manual + misc + movie + music + mvcollection + mvreference + mvproceedings + mvbook + online + patent + performance + periodical + proceedings + reference + report + review + set + software + standard + suppbook + suppcollection + suppperiodical + thesis + unpublished + video + xdata + + + sortyear + volume + volumes + abstract + addendum + annotation + booksubtitle + booktitle + booktitleaddon + chapter + edition + eid + entrysubtype + eprintclass + eprinttype + eventtitle + eventtitleaddon + gender + howpublished + indexsorttitle + indextitle + isan + isbn + ismn + isrn + issn + issue + issuesubtitle + issuetitle + issuetitleaddon + iswc + journalsubtitle + journaltitle + journaltitleaddon + label + langid + langidopts + library + mainsubtitle + maintitle + maintitleaddon + nameaddon + note + number + origtitle + pagetotal + part + relatedstring + relatedtype + reprinttitle + series + shorthandintro + subtitle + title + titleaddon + usera + userb + userc + userd + usere + userf + venue + version + shorthand + shortjournal + shortseries + shorttitle + sorttitle + sortshorthand + sortkey + presort + institution + lista + listb + listc + listd + liste + listf + location + organization + origlocation + origpublisher + publisher + afterword + annotator + author + bookauthor + commentator + editor + editora + editorb + editorc + foreword + holder + introduction + namea + nameb + namec + translator + shortauthor + shorteditor + sortname + authortype + editoratype + editorbtype + editorctype + editortype + bookpagination + nameatype + namebtype + namectype + pagination + pubstate + type + language + origlanguage + crossref + xref + date + endyear + year + month + day + hour + minute + second + timezone + yeardivision + endmonth + endday + endhour + endminute + endsecond + endtimezone + endyeardivision + eventdate + eventendyear + eventyear + eventmonth + eventday + eventhour + eventminute + eventsecond + eventtimezone + eventyeardivision + eventendmonth + eventendday + eventendhour + eventendminute + eventendsecond + eventendtimezone + eventendyeardivision + origdate + origendyear + origyear + origmonth + origday + orighour + origminute + origsecond + origtimezone + origyeardivision + origendmonth + origendday + origendhour + origendminute + origendsecond + origendtimezone + origendyeardivision + urldate + urlendyear + urlyear + urlmonth + urlday + urlhour + urlminute + urlsecond + urltimezone + urlyeardivision + urlendmonth + urlendday + urlendhour + urlendminute + urlendsecond + urlendtimezone + urlendyeardivision + doi + eprint + file + verba + verbb + verbc + url + xdata + ids + entryset + related + keywords + options + relatedoptions + pages + execute + + + abstract + annotation + authortype + bookpagination + crossref + day + doi + eprint + eprintclass + eprinttype + endday + endhour + endminute + endmonth + endsecond + endtimezone + endyear + endyeardivision + entryset + entrysubtype + execute + file + gender + hour + ids + indextitle + indexsorttitle + isan + ismn + iswc + keywords + label + langid + langidopts + library + lista + listb + listc + listd + liste + listf + minute + month + namea + nameb + namec + nameatype + namebtype + namectype + nameaddon + options + origday + origendday + origendhour + origendminute + origendmonth + origendsecond + origendtimezone + origendyear + origendyeardivision + orighour + origminute + origmonth + origsecond + origtimezone + origyear + origyeardivision + origlocation + origpublisher + origtitle + pagination + presort + related + relatedoptions + relatedstring + relatedtype + second + shortauthor + shorteditor + shorthand + shorthandintro + shortjournal + shortseries + shorttitle + sortkey + sortname + sortshorthand + sorttitle + sortyear + timezone + url + urlday + urlendday + urlendhour + urlendminute + urlendmonth + urlendsecond + urlendtimezone + urlendyear + urlhour + urlminute + urlmonth + urlsecond + urltimezone + urlyear + usera + userb + userc + userd + usere + userf + verba + verbb + verbc + xdata + xref + year + yeardivision + + + set + entryset + + + article + addendum + annotator + author + commentator + editor + editora + editorb + editorc + editortype + editoratype + editorbtype + editorctype + eid + issn + issue + issuetitle + issuesubtitle + issuetitleaddon + journalsubtitle + journaltitle + journaltitleaddon + language + note + number + origlanguage + pages + pubstate + series + subtitle + title + titleaddon + translator + version + volume + + + bibnote + note + + + book + author + addendum + afterword + annotator + chapter + commentator + edition + editor + editora + editorb + editorc + editortype + editoratype + editorbtype + editorctype + eid + foreword + introduction + isbn + language + location + maintitle + maintitleaddon + mainsubtitle + note + number + origlanguage + pages + pagetotal + part + publisher + pubstate + series + subtitle + title + titleaddon + translator + volume + volumes + + + mvbook + addendum + afterword + annotator + author + commentator + edition + editor + editora + editorb + editorc + editortype + editoratype + editorbtype + editorctype + foreword + introduction + isbn + language + location + note + number + origlanguage + pagetotal + publisher + pubstate + series + subtitle + title + titleaddon + translator + volume + volumes + + + inbook + bookinbook + suppbook + addendum + afterword + annotator + author + booktitle + bookauthor + booksubtitle + booktitleaddon + chapter + commentator + edition + editor + editora + editorb + editorc + editortype + editoratype + editorbtype + editorctype + eid + foreword + introduction + isbn + language + location + mainsubtitle + maintitle + maintitleaddon + note + number + origlanguage + part + publisher + pages + pubstate + series + subtitle + title + titleaddon + translator + volume + volumes + + + booklet + addendum + author + chapter + editor + editortype + eid + howpublished + language + location + note + pages + pagetotal + pubstate + subtitle + title + titleaddon + type + + + collection + reference + addendum + afterword + annotator + chapter + commentator + edition + editor + editora + editorb + editorc + editortype + editoratype + editorbtype + editorctype + eid + foreword + introduction + isbn + language + location + mainsubtitle + maintitle + maintitleaddon + note + number + origlanguage + pages + pagetotal + part + publisher + pubstate + series + subtitle + title + titleaddon + translator + volume + volumes + + + mvcollection + mvreference + addendum + afterword + annotator + author + commentator + edition + editor + editora + editorb + editorc + editortype + editoratype + editorbtype + editorctype + foreword + introduction + isbn + language + location + note + number + origlanguage + publisher + pubstate + subtitle + title + titleaddon + translator + volume + volumes + + + incollection + suppcollection + inreference + addendum + afterword + annotator + author + booksubtitle + booktitle + booktitleaddon + chapter + commentator + edition + editor + editora + editorb + editorc + editortype + editoratype + editorbtype + editorctype + eid + foreword + introduction + isbn + language + location + mainsubtitle + maintitle + maintitleaddon + note + number + origlanguage + pages + part + publisher + pubstate + series + subtitle + title + titleaddon + translator + volume + volumes + + + dataset + addendum + author + edition + editor + editortype + language + location + note + number + organization + publisher + pubstate + series + subtitle + title + titleaddon + type + version + + + manual + addendum + author + chapter + edition + editor + editortype + eid + isbn + language + location + note + number + organization + pages + pagetotal + publisher + pubstate + series + subtitle + title + titleaddon + type + version + + + misc + software + addendum + author + editor + editortype + howpublished + language + location + note + organization + pubstate + subtitle + title + titleaddon + type + version + + + online + addendum + author + editor + editortype + language + note + organization + pubstate + subtitle + title + titleaddon + version + + + patent + addendum + author + holder + location + note + number + pubstate + subtitle + title + titleaddon + type + version + + + periodical + addendum + editor + editora + editorb + editorc + editortype + editoratype + editorbtype + editorctype + issn + issue + issuesubtitle + issuetitle + issuetitleaddon + language + note + number + pubstate + series + subtitle + title + titleaddon + volume + yeardivision + + + mvproceedings + addendum + editor + editortype + eventday + eventendday + eventendhour + eventendminute + eventendmonth + eventendsecond + eventendtimezone + eventendyear + eventendyeardivision + eventhour + eventminute + eventmonth + eventsecond + eventtimezone + eventyear + eventyeardivision + eventtitle + eventtitleaddon + isbn + language + location + note + number + organization + pagetotal + publisher + pubstate + series + subtitle + title + titleaddon + venue + volumes + + + proceedings + addendum + chapter + editor + editortype + eid + eventday + eventendday + eventendhour + eventendminute + eventendmonth + eventendsecond + eventendtimezone + eventendyear + eventendyeardivision + eventhour + eventminute + eventmonth + eventsecond + eventtimezone + eventyear + eventyeardivision + eventtitle + eventtitleaddon + isbn + language + location + mainsubtitle + maintitle + maintitleaddon + note + number + organization + pages + pagetotal + part + publisher + pubstate + series + subtitle + title + titleaddon + venue + volume + volumes + + + inproceedings + addendum + author + booksubtitle + booktitle + booktitleaddon + chapter + editor + editortype + eid + eventday + eventendday + eventendhour + eventendminute + eventendmonth + eventendsecond + eventendtimezone + eventendyear + eventendyeardivision + eventhour + eventminute + eventmonth + eventsecond + eventtimezone + eventyear + eventyeardivision + eventtitle + eventtitleaddon + isbn + language + location + mainsubtitle + maintitle + maintitleaddon + note + number + organization + pages + part + publisher + pubstate + series + subtitle + title + titleaddon + venue + volume + volumes + + + report + addendum + author + chapter + eid + institution + isrn + language + location + note + number + pages + pagetotal + pubstate + subtitle + title + titleaddon + type + version + + + thesis + addendum + author + chapter + eid + institution + language + location + note + pages + pagetotal + pubstate + subtitle + title + titleaddon + type + + + unpublished + addendum + author + eventday + eventendday + eventendhour + eventendminute + eventendmonth + eventendsecond + eventendtimezone + eventendyear + eventendyeardivision + eventhour + eventminute + eventmonth + eventsecond + eventtimezone + eventyear + eventyeardivision + eventtitle + eventtitleaddon + howpublished + language + location + note + pubstate + subtitle + title + titleaddon + type + venue + + + abstract + addendum + afterword + annotator + author + bookauthor + booksubtitle + booktitle + booktitleaddon + chapter + commentator + editor + editora + editorb + editorc + foreword + holder + institution + introduction + issuesubtitle + issuetitle + issuetitleaddon + journalsubtitle + journaltitle + journaltitleaddon + location + mainsubtitle + maintitle + maintitleaddon + nameaddon + note + organization + origlanguage + origlocation + origpublisher + origtitle + part + publisher + relatedstring + series + shortauthor + shorteditor + shorthand + shortjournal + shortseries + shorttitle + sortname + sortshorthand + sorttitle + subtitle + title + titleaddon + translator + venue + + + article + book + inbook + bookinbook + suppbook + booklet + collection + incollection + suppcollection + manual + misc + mvbook + mvcollection + online + patent + periodical + suppperiodical + proceedings + inproceedings + reference + inreference + report + set + thesis + unpublished + + + date + year + + + + + set + + entryset + + + + article + + author + journaltitle + title + + + + book + mvbook + + author + title + + + + inbook + bookinbook + suppbook + + author + title + booktitle + + + + booklet + + + author + editor + + title + + + + collection + reference + mvcollection + mvreference + + editor + title + + + + incollection + suppcollection + inreference + + author + editor + title + booktitle + + + + dataset + + title + + + + manual + + title + + + + misc + software + + title + + + + online + + title + + url + doi + eprint + + + + + patent + + author + title + number + + + + periodical + + editor + title + + + + proceedings + mvproceedings + + title + + + + inproceedings + + author + title + booktitle + + + + report + + author + title + type + institution + + + + thesis + + author + title + type + institution + + + + unpublished + + author + title + + + + + isbn + + + issn + + + ismn + + + gender + + + + + + + bibliography.bib + + + scaramuzzaVisualOdometryTutorial2011 + durrant-whyteSimultaneousLocalizationMapping2006 + davisonMonoSLAMRealtimeSingle2007 + kerlDenseVisualSLAM2013 + newcombeKinectfusionRealtimeDense2011 + rusinkiewiczEfficientVariantsICP2001 + mildenhallNeRFRepresentingScenes2022 + sandstromPointslamDenseNeural2023 + sucarImapImplicitMapping2021 + zhuNiceslamNeuralImplicit2022 + garbinFastnerfHighfidelityNeural2021 + kerbl3DGaussianSplatting2023 + pengRTGSLAMRealtime3D2024 + haRGBDGSICPSLAM2024 + yugayGaussianSLAMPhotorealisticDense2024 + mur-artalOrbslam2OpensourceSlam2017 + camposOrbslam3AccurateOpensource2021 + gauglitzEvaluationInterestPoint2011 + engelDirectSparseOdometry2017 + kerlDenseVisualSLAM2013 + kerlRobustOdometryEstimation2013 + newcombeDTAMDenseTracking2011 + whelanElasticFusionRealtimeDense2016 + kerlRobustOdometryEstimation2013 + mildenhallNeRFRepresentingScenes2022 + yen-chenInerfInvertingNeural2021 + mullerInstantNeuralGraphics2022 + yuPlenoctreesRealtimeRendering2021 + fridovich-keilPlenoxelsRadianceFields2022 + kerbl3DGaussianSplatting2023 + keethaSplaTAMSplatTrack2024 + huCGSLAMEfficientDense2024 + pengRTGSLAMRealtime3D2024 + haRGBDGSICPSLAM2024 + beslMethodRegistration3shapes1992 + pengRTGSLAMRealtime3D2024 + segalGeneralizedicp2009a + yugayGaussianSLAMPhotorealisticDense2024 + parkColoredPointCloud2017 + steinbruckerRealtimeVisualOdometry2011 + pomerleauComparingICPVariants2013 + kuipersQuaternionsRotationSequences1999 + mildenhallNeRFRepresentingScenes2022 + kerbl3DGaussianSplatting2023 + kerbl3DGaussianSplatting2023 + kerbl3DGaussianSplatting2023 + zwickerEWASplatting2002 + kerbl3DGaussianSplatting2023 + kuipersQuaternionsRotationSequences1999 + kanopoulosDesignImageEdge1988 + kingmaAdamMethodStochastic2014 + straubReplicaDatasetDigital2019 + sturmBenchmarkEvaluationRGBD2012 + sucarImapImplicitMapping2021 + sturmBenchmarkEvaluationRGBD2012 + pengRTGSLAMRealtime3D2024 + haRGBDGSICPSLAM2024 + yugayGaussianSLAMPhotorealisticDense2024 + straubReplicaDatasetDigital2019 + straubReplicaDatasetDigital2019 + pengRTGSLAMRealtime3D2024 + haRGBDGSICPSLAM2024 + yugayGaussianSLAMPhotorealisticDense2024 + yugayGaussianSLAMPhotorealisticDense2024 + pengRTGSLAMRealtime3D2024 + straubReplicaDatasetDigital2019 + straubReplicaDatasetDigital2019 + pengRTGSLAMRealtime3D2024 + haRGBDGSICPSLAM2024 + yugayGaussianSLAMPhotorealisticDense2024 + yugayGaussianSLAMPhotorealisticDense2024 + pengRTGSLAMRealtime3D2024 + haRGBDGSICPSLAM2024 + sturmBenchmarkEvaluationRGBD2012 + sturmBenchmarkEvaluationRGBD2012 + pengRTGSLAMRealtime3D2024 + haRGBDGSICPSLAM2024 + yugayGaussianSLAMPhotorealisticDense2024 + yugayGaussianSLAMPhotorealisticDense2024 + haRGBDGSICPSLAM2024 + yugayGaussianSLAMPhotorealisticDense2024 + pengRTGSLAMRealtime3D2024 + sturmBenchmarkEvaluationRGBD2012 + sturmBenchmarkEvaluationRGBD2012 + pengRTGSLAMRealtime3D2024 + haRGBDGSICPSLAM2024 + yugayGaussianSLAMPhotorealisticDense2024 + sturmBenchmarkEvaluationRGBD2012 + sturmBenchmarkEvaluationRGBD2012 + pengRTGSLAMRealtime3D2024 + haRGBDGSICPSLAM2024 + yugayGaussianSLAMPhotorealisticDense2024 + yugayGaussianSLAMPhotorealisticDense2024 + sturmBenchmarkEvaluationRGBD2012 + sturmBenchmarkEvaluationRGBD2012 + sturmBenchmarkEvaluationRGBD2012 + sturmBenchmarkEvaluationRGBD2012 + sturmBenchmarkEvaluationRGBD2012 + + + + + citeorder + + + intciteorder + + + + + + diff --git a/docs/submission/main.blg b/docs/submission/main.blg new file mode 100644 index 0000000..dcd99a9 --- /dev/null +++ b/docs/submission/main.blg @@ -0,0 +1,15 @@ +[0] Config.pm:307> INFO - This is Biber 2.19 +[0] Config.pm:310> INFO - Logfile is 'main.blg' +[99] biber-MSWIN64:340> INFO - === +[138] Biber.pm:419> INFO - Reading 'main.bcf' +[248] Biber.pm:979> INFO - Found 39 citekeys in bib section 0 +[262] Biber.pm:4419> INFO - Processing section 0 +[274] Biber.pm:4610> INFO - Looking for bibtex file 'bibliography.bib' for section 0 +[360] bibtex.pm:1713> INFO - LaTeX decoding ... +[563] bibtex.pm:1519> INFO - Found BibTeX data source 'bibliography.bib' +[882] UCollate.pm:68> INFO - Overriding locale 'en-US' defaults 'normalization = NFD' with 'normalization = prenormalized' +[883] UCollate.pm:68> INFO - Overriding locale 'en-US' defaults 'variable = shifted' with 'variable = non-ignorable' +[883] Biber.pm:4239> INFO - Sorting list 'none/global//global/global' of type 'entry' with template 'none' and locale 'en-US' +[883] Biber.pm:4245> INFO - No sort tailoring available for locale 'en-US' +[912] bbl.pm:660> INFO - Writing 'main.bbl' with encoding 'UTF-8' +[950] bbl.pm:763> INFO - Output to main.bbl diff --git a/docs/submission/main.out b/docs/submission/main.out new file mode 100644 index 0000000..9d13ee8 --- /dev/null +++ b/docs/submission/main.out @@ -0,0 +1,16 @@ +\BOOKMARK [1][-]{section.1}{\376\377\000I\000n\000t\000r\000o\000d\000u\000c\000t\000i\000o\000n}{}% 1 +\BOOKMARK [1][-]{section.2}{\376\377\000R\000e\000l\000a\000t\000e\000d\000\040\000W\000o\000r\000k}{}% 2 +\BOOKMARK [2][-]{subsection.2.1}{\376\377\000C\000l\000a\000s\000s\000i\000c\000a\000l\000\040\000R\000G\000B\000-\000D\000\040\000L\000o\000c\000a\000l\000i\000z\000a\000t\000i\000o\000n}{section.2}% 3 +\BOOKMARK [2][-]{subsection.2.2}{\376\377\000N\000e\000R\000F\000-\000B\000a\000s\000e\000d\000\040\000L\000o\000c\000a\000l\000i\000z\000a\000t\000i\000o\000n}{section.2}% 4 +\BOOKMARK [2][-]{subsection.2.3}{\376\377\000G\000a\000u\000s\000s\000i\000a\000n\000-\000B\000a\000s\000e\000d\000\040\000L\000o\000c\000a\000l\000i\000z\000a\000t\000i\000o\000n}{section.2}% 5 +\BOOKMARK [1][-]{section.3}{\376\377\000M\000e\000t\000h\000o\000d}{}% 6 +\BOOKMARK [2][-]{subsection.3.1}{\376\377\000S\000c\000e\000n\000e\000\040\000R\000e\000p\000r\000e\000s\000e\000n\000t\000a\000t\000i\000o\000n}{section.3}% 7 +\BOOKMARK [2][-]{subsection.3.2}{\376\377\000D\000e\000p\000t\000h\000\040\000R\000e\000n\000d\000e\000r\000i\000n\000g}{section.3}% 8 +\BOOKMARK [2][-]{subsection.3.3}{\376\377\000L\000o\000c\000a\000l\000i\000z\000a\000t\000i\000o\000n\000\040\000a\000s\000\040\000I\000m\000a\000g\000e\000\040\000A\000l\000i\000g\000n\000m\000e\000n\000t}{section.3}% 9 +\BOOKMARK [2][-]{subsection.3.4}{\376\377\000P\000i\000p\000e\000l\000i\000n\000e}{section.3}% 10 +\BOOKMARK [1][-]{section.4}{\376\377\000E\000v\000a\000l\000u\000a\000t\000i\000o\000n}{}% 11 +\BOOKMARK [2][-]{subsection.4.1}{\376\377\000E\000x\000p\000e\000r\000i\000m\000e\000n\000t\000a\000l\000\040\000S\000e\000t\000u\000p}{section.4}% 12 +\BOOKMARK [2][-]{subsection.4.2}{\376\377\000L\000o\000c\000a\000l\000i\000z\000a\000t\000i\000o\000n\000\040\000E\000v\000a\000l\000u\000a\000t\000i\000o\000n}{section.4}% 13 +\BOOKMARK [2][-]{subsection.4.3}{\376\377\000D\000i\000s\000c\000u\000s\000s\000i\000o\000n}{section.4}% 14 +\BOOKMARK [2][-]{subsection.4.4}{\376\377\000L\000i\000m\000i\000t\000a\000t\000i\000o\000n\000s}{section.4}% 15 +\BOOKMARK [1][-]{section.5}{\376\377\000C\000o\000n\000c\000l\000u\000s\000i\000o\000n}{}% 16 diff --git a/docs/submission/main.pdf b/docs/submission/main.pdf index af473e7..878e27d 100644 Binary files a/docs/submission/main.pdf and b/docs/submission/main.pdf differ diff --git a/docs/submission/main.run.xml b/docs/submission/main.run.xml new file mode 100644 index 0000000..a476a7a --- /dev/null +++ b/docs/submission/main.run.xml @@ -0,0 +1,89 @@ + + + + + + + + + + + + + + + + + + + + + + + + +]> + + + latex + + main.bcf + + + main.bbl + + + blx-dm.def + blx-unicode.def + blx-compat.def + biblatex.def + standard.bbx + numeric.bbx + numeric-comp.bbx + ieee.bbx + numeric-verb.cbx + ieee.cbx + biblatex.cfg + english.lbx + + + + biber + + biber + main + + + main.bcf + + + main.bbl + + + main.bbl + + + main.bcf + + + bibliography.bib + + + diff --git a/docs/submission/main.tex b/docs/submission/main.tex index 2091b24..01f1c47 100644 --- a/docs/submission/main.tex +++ b/docs/submission/main.tex @@ -1,18 +1,18 @@ -\documentclass[twocolumn]{article} % 启用双栏排版 +\documentclass[twocolumn]{article} \usepackage{mathtools} \usepackage{unicode-math} -\usepackage{arxiv} % 特定于arxiv的样式包，用于格式设置 +\usepackage{arxiv} \usepackage{fontspec} -\usepackage{url} % 用于类型设置URL -\usepackage{booktabs} % 创建专业质量的表格 -\usepackage{amsfonts} % 黑板数学符号 -\usepackage{nicefrac} % 紧凑的分数符号 -\usepackage{microtype} % 微排版 -\usepackage{lipsum} % 生成填充文本 -\usepackage{graphicx} % 图形包 -\usepackage{doi} % 处理DOI -\usepackage{titlesec} % 调整节标题的间距和格式 +\usepackage{url} +\usepackage{booktabs} +\usepackage{amsfonts} +\usepackage{nicefrac} +\usepackage{microtype} +\usepackage{lipsum} +\usepackage{graphicx} +\usepackage{doi} +\usepackage{titlesec} \usepackage{setspace} \setstretch{1.1} \usepackage{xcolor} @@ -31,13 +31,13 @@ \captionsetup{font=small,labelfont=bf} - -% 标题设置 \title{\large\bfseries\textit{GSplatLoc} : Ultra-Precise Camera Localization via 3D Gaussian Splatting} \usepackage{titling} + +% https://www.overleaf.com/learn/latex/Hyperlinks \usepackage{hyperref} \hypersetup{ @@ -51,14 +51,13 @@ } \urlstyle{same} -% 自定义作者信息命令 + \newcommand{\authorinfo}[4]{% \begin{tabular}[t]{c} \href{#1}{\includegraphics[scale=0.06]{orcid.pdf}\hspace{1mm}\textbf{#2}}#4\\ #3\\ \end{tabular}% } -% 重定义 \and 命令以添加适当的间距 \renewcommand{\and}{\hspace{2em}} % 作者设置 @@ -67,14 +66,12 @@ Zeller}{Southeast University Chengxian College\\Nanjing, China}{}\and% } - % main \begin{document} \twocolumn[ \begin{@twocolumnfalse} \maketitle - % \correspondingauthor \begin{abstract} We present \textbf{GSplatLoc}, a camera localization method that leverages the differentiable rendering capabilities of 3D Gaussian @@ -98,37 +95,37 @@ \section{Introduction}\label{introduction} -Visual localization\cite{scaramuzzaVisualOdometryTutorial2011}, +Visual localization\autocite{scaramuzzaVisualOdometryTutorial2011}, specifically the task of estimating camera position and orientation (pose estimation) for a given image within a known scene, is a fundamental challenge in computer vision. Accurate pose estimation is crucial for applications like autonomous robotics (e.g., self-driving cars), as well as Augmented and Virtual Reality systems. Although Visual Simultaneous Localization and Mapping (Visual -SLAM)\cite{durrant-whyteSimultaneousLocalizationMapping2006,davisonMonoSLAMRealtimeSingle2007}combines +SLAM)\autocite{durrant-whyteSimultaneousLocalizationMapping2006,davisonMonoSLAMRealtimeSingle2007}combines both mapping and pose estimation, this paper focuses specifically on the localization component, which is essential for real-time tracking in dynamic environments. -Traditional SLAM systems \cite{kerlDenseVisualSLAM2013} have +Traditional SLAM systems \autocite{kerlDenseVisualSLAM2013} have demonstrated accurate pose estimation across diverse environments. However, their underlying 3D representations (e.g., point clouds, meshes, and surfels) exhibit -limitations\cite{newcombeKinectfusionRealtimeDense2011,rusinkiewiczEfficientVariantsICP2001} +limitations\autocite{newcombeKinectfusionRealtimeDense2011,rusinkiewiczEfficientVariantsICP2001} in flexibility for tasks like photorealistic scene exploration and fine-grained map updates. Recent methods utilizing Neural Radiance -Fields (NeRF) \cite{mildenhallNeRFRepresentingScenes2022} for +Fields (NeRF) \autocite{mildenhallNeRFRepresentingScenes2022} for surface reconstruction and view rendering have inspired novel SLAM -approaches \cite{sandstromPointslamDenseNeural2023}, which show -promising\cite{sucarImapImplicitMapping2021,zhuNiceslamNeuralImplicit2022} +approaches \autocite{sandstromPointslamDenseNeural2023}, which show +promising\autocite{sucarImapImplicitMapping2021,zhuNiceslamNeuralImplicit2022} results in tracking and scene modeling. Despite these -advances\cite{garbinFastnerfHighfidelityNeural2021}, existing +advances\autocite{garbinFastnerfHighfidelityNeural2021}, existing NeRF-based methods rely on computationally expensive volume rendering pipelines, limiting their ability to perform real-time \textbf{pose estimation} effectively. The development of \textbf{3D Gaussian Splatting} -\cite{kerbl3DGaussianSplatting2023} for efficient novel view +\autocite{kerbl3DGaussianSplatting2023} for efficient novel view synthesis presents a promising solution to these limitations. Its rasterization-based rendering pipeline enables faster image-level rendering, making it more suitable for real-time applications. However, @@ -137,10 +134,10 @@ \section{Introduction}\label{introduction} and a lack of explicit multi-view constraints. Current SLAM methods using 3D Gaussian Splatting, such as RTG-SLAM -\cite{pengRTGSLAMRealtime3D2024} and GS-ICP-SLAM -\cite{haRGBDGSICPSLAM2024}, rely primarily on ICP-based techniques +\autocite{pengRTGSLAMRealtime3D2024} and GS-ICP-SLAM +\autocite{haRGBDGSICPSLAM2024}, rely primarily on ICP-based techniques for pose estimation. Other approaches, like Gaussian-SLAM -\cite{yugayGaussianSLAMPhotorealisticDense2024}, adapt traditional +\autocite{yugayGaussianSLAMPhotorealisticDense2024}, adapt traditional RGB-D odometry methods. While these methods have shown potential, they often do not fully exploit the differentiable nature of the Gaussian Splatting representation, particularly for real-time and efficient @@ -148,39 +145,27 @@ \section{Introduction}\label{introduction} In this paper, we introduce \textbf{GSplatLoc}, a novel camera localization method that leverages the differentiable properties of 3D -Gaussian Splatting specifically for efficient and accurate \textbf{pose -estimation}. Rather than addressing the full SLAM pipeline, our approach -is designed to focus solely on the localization aspect, allowing for -more efficient use of the scene representation and camera pose -estimation. By developing a fully differentiable pipeline, GSplatLoc can -be seamlessly integrated into existing Gaussian Splatting SLAM -frameworks or other deep learning tasks focused on localization. - -Our main contributions are as follows: - -\begin{enumerate} -\def\labelenumi{\arabic{enumi}.} -\item - We present a GPU-accelerated framework for real-time camera - localization, based on a comprehensive theoretical analysis of camera - pose derivatives in 3D Gaussian Splatting. -\item - We propose a novel optimization approach that focuses on camera pose - estimation given a 3D Gaussian scene, fully exploiting the - differentiable nature of the rendering process. -\item - We demonstrate the effectiveness of our method through extensive - experiments, showing competitive or superior pose estimation results - compared to state-of-the-art SLAM approaches utilizing advanced scene - representations. -\end{enumerate} - -By focusing specifically on the challenges of localization in Gaussian -Splatting-based scenes, GSplatLoc opens new avenues for high-precision -\textbf{camera pose estimation} in complex environments. Our work -contributes to the ongoing advancement of visual localization systems, -pushing the boundaries of accuracy and real-time performance in 3D scene -understanding and navigation. +Gaussian Splatting for efficient and accurate pose estimation. By +focusing solely on the localization aspect rather than the full SLAM +pipeline, GSplatLoc allows for more efficient utilization of the scene +representation and camera pose estimation, seamlessly integrating into +existing Gaussian Splatting SLAM frameworks or other deep learning tasks +focused on localization. + +Our main contributions include presenting a GPU-accelerated framework +for real-time camera localization, based on a comprehensive theoretical +analysis of camera pose derivatives in 3D Gaussian Splatting; proposing +a novel optimization approach that fully exploits the differentiable +nature of the rendering process for camera pose estimation given a 3D +Gaussian scene; and demonstrating the effectiveness of our method +through extensive experiments, showing competitive or superior pose +estimation results compared to state-of-the-art SLAM approaches +utilizing advanced scene representations. By specifically addressing the +challenges of localization in Gaussian Splatting-based scenes, GSplatLoc +opens new avenues for high-precision camera pose estimation in complex +environments, contributing to the ongoing advancement of visual +localization systems and pushing the boundaries of accuracy and +real-time performance in 3D scene understanding and navigation. \section{Related Work}\label{related-work} @@ -204,9 +189,9 @@ \subsection{Classical RGB-D \textbf{Feature-Based Methods} involve extracting and matching keypoints across frames to estimate camera motion. Notable systems such as -ORB-SLAM2 \cite{mur-artalOrbslam2OpensourceSlam2017} , ORB-SLAM3 -\cite{camposOrbslam3AccurateOpensource2021} and -\cite{gauglitzEvaluationInterestPoint2011} rely on sparse feature +ORB-SLAM2 \autocite{mur-artalOrbslam2OpensourceSlam2017} , ORB-SLAM3 +\autocite{camposOrbslam3AccurateOpensource2021} and +\autocite{gauglitzEvaluationInterestPoint2011} rely on sparse feature descriptors like ORB features. These systems have demonstrated robust performance in various environments, benefiting from the maturity of feature detection and matching algorithms. However, their reliance on @@ -216,12 +201,12 @@ \subsection{Classical RGB-D pose estimation, making them susceptible to lighting changes and appearance variations. -\textbf{Direct Methods}\cite{engelDirectSparseOdometry2017} estimate +\textbf{Direct Methods}\autocite{engelDirectSparseOdometry2017} estimate camera motion by minimizing the photometric error between consecutive frames, utilizing all available pixel information. Methods such as Dense Visual Odometry (DVO) -\cite{kerlDenseVisualSLAM2013,kerlRobustOdometryEstimation2013} and -DTAM\cite{newcombeDTAMDenseTracking2011} incorporate depth data to +\autocite{kerlDenseVisualSLAM2013,kerlRobustOdometryEstimation2013} and +DTAM\autocite{newcombeDTAMDenseTracking2011} incorporate depth data to enhance pose estimation accuracy. These methods can achieve high precision in well-lit, textured environments but are sensitive to illumination changes and require good initialization to avoid local @@ -232,10 +217,10 @@ \subsection{Classical RGB-D \textbf{Hybrid Approaches} combine the strengths of feature-based and direct methods. ElasticFusion -\cite{whelanElasticFusionRealtimeDense2016} integrates surfel-based +\autocite{whelanElasticFusionRealtimeDense2016} integrates surfel-based mapping with real-time camera tracking, using both photometric and geometric information. DVO-SLAM -\cite{kerlRobustOdometryEstimation2013} combines geometric and +\autocite{kerlRobustOdometryEstimation2013} combines geometric and photometric alignment for improved robustness. However, these methods often involve complex pipelines and can be computationally intensive due to dense map representations and intricate data association processes. @@ -251,7 +236,7 @@ \subsection{Classical RGB-D \subsection{NeRF-Based Localization}\label{nerf-based-localization} The advent of Neural Radiance Fields (NeRF) -\cite{mildenhallNeRFRepresentingScenes2022} has revolutionized novel +\autocite{mildenhallNeRFRepresentingScenes2022} has revolutionized novel view synthesis by representing scenes as continuous volumetric functions learned from images. NeRF has inspired new approaches to camera localization by leveraging its differentiable rendering capabilities. @@ -259,7 +244,7 @@ \subsection{NeRF-Based Localization}\label{nerf-based-localization} \textbf{Pose Estimation with NeRF} involves inverting a pre-trained NeRF model to recover camera poses by minimizing the photometric error between rendered images and observed images. iNeRF -\cite{yen-chenInerfInvertingNeural2021} formulates pose estimation +\autocite{yen-chenInerfInvertingNeural2021} formulates pose estimation as an optimization problem, using gradient-based methods to refine camera parameters. While iNeRF achieves impressive accuracy, it suffers from high computational costs due to the per-pixel ray marching required @@ -268,10 +253,10 @@ \subsection{NeRF-Based Localization}\label{nerf-based-localization} \textbf{Accelerated NeRF Variants} aim to address computational inefficiency by introducing explicit data structures. Instant-NGP -\cite{mullerInstantNeuralGraphics2022} uses hash maps to accelerate +\autocite{mullerInstantNeuralGraphics2022} uses hash maps to accelerate training and rendering, achieving interactive frame rates. PlenOctrees -\cite{yuPlenoctreesRealtimeRendering2021} and Plenoxels -\cite{fridovich-keilPlenoxelsRadianceFields2022} employ sparse voxel +\autocite{yuPlenoctreesRealtimeRendering2021} and Plenoxels +\autocite{fridovich-keilPlenoxelsRadianceFields2022} employ sparse voxel grids to represent the scene, significantly reducing computation time. However, even with these optimizations, rendering speeds may still not meet the demands of real-time localization in dynamic environments. @@ -288,16 +273,16 @@ \subsection{Gaussian-Based Recent advancements in scene representation have introduced 3D Gaussian splatting as an efficient alternative to NeRF. \textbf{3D Gaussian -Splatting} \cite{kerbl3DGaussianSplatting2023} represents scenes +Splatting} \autocite{kerbl3DGaussianSplatting2023} represents scenes using a set of 3D Gaussian primitives and employs rasterization-based rendering, offering significant computational advantages over volumetric rendering. \textbf{Gaussian Splatting in Localization} has been explored in methods -such as SplaTAM \cite{keethaSplaTAMSplatTrack2024}, CG-SLAM -\cite{huCGSLAMEfficientDense2024}, RTG-SLAM -\cite{pengRTGSLAMRealtime3D2024}, and GS-ICP-SLAM -\cite{haRGBDGSICPSLAM2024}. SplaTAM introduces a SLAM system that +such as SplaTAM \autocite{keethaSplaTAMSplatTrack2024}, CG-SLAM +\autocite{huCGSLAMEfficientDense2024}, RTG-SLAM +\autocite{pengRTGSLAMRealtime3D2024}, and GS-ICP-SLAM +\autocite{haRGBDGSICPSLAM2024}. SplaTAM introduces a SLAM system that uses gradient-based optimization to refine both the map and camera poses, utilizing RGB-D data and 3D Gaussians for dense mapping. CG-SLAM focuses on an uncertainty-aware 3D Gaussian field to improve tracking @@ -305,24 +290,24 @@ \subsection{Gaussian-Based Pose estimation approaches in these methods often rely on traditional point cloud registration techniques, such as Iterative Closest Point -(ICP) algorithms \cite{beslMethodRegistration3shapes1992}. -\textbf{RTG-SLAM} employs ICP for pose estimation within a 3D Gaussian -splatting framework, demonstrating real-time performance in 3D -reconstruction tasks. Similarly, \textbf{GS-ICP-SLAM} utilizes -Generalized ICP \cite{segalGeneralizedicp2009a} for alignment, -effectively handling the variability in point cloud density and -improving robustness. +(ICP) algorithms \autocite{beslMethodRegistration3shapes1992}. +\textbf{RTG-SLAM}\autocite{pengRTGSLAMRealtime3D2024} employs ICP for +pose estimation within a 3D Gaussian splatting framework, demonstrating +real-time performance in 3D reconstruction tasks. Similarly, +\textbf{GS-ICP-SLAM} utilizes Generalized ICP +\autocite{segalGeneralizedicp2009a} for alignment, effectively handling +the variability in point cloud density and improving robustness. \textbf{Gaussian-SLAM} -\cite{yugayGaussianSLAMPhotorealisticDense2024} adapts traditional +\autocite{yugayGaussianSLAMPhotorealisticDense2024} adapts traditional RGB-D odometry methods, combining colored point cloud alignment -\cite{parkColoredPointCloud2017} with an energy-based visual -odometry approach \cite{steinbruckerRealtimeVisualOdometry2011}. +\autocite{parkColoredPointCloud2017} with an energy-based visual +odometry approach \autocite{steinbruckerRealtimeVisualOdometry2011}. These methods integrate ICP-based techniques within Gaussian-based representations to estimate camera poses. While effective in certain scenarios, the reliance on ICP-based methods -introduces limitations\cite{pomerleauComparingICPVariants2013}. ICP +introduces limitations\autocite{pomerleauComparingICPVariants2013}. ICP algorithms require good initial alignment and can be sensitive to local minima, often necessitating careful initialization to ensure convergence. Additionally, ICP can be computationally intensive, @@ -356,7 +341,7 @@ \subsection{Gaussian-Based localization from mapping, we simplify the optimization process, making it more suitable for real-time applications. Additionally, using quaternions for rotation parameterization -\cite{kuipersQuaternionsRotationSequences1999} and careful +\autocite{kuipersQuaternionsRotationSequences1999} and careful initialization strategies improves the stability and convergence of the optimization, addressing challenges associated with sensor noise and incomplete data. @@ -393,13 +378,13 @@ \section{Method}\label{method} rendering quality and computational efficiency, hindering their ability to provide photorealistic scene exploration and fine-grained map updates. Neural Radiance Fields (NeRF) -\cite{mildenhallNeRFRepresentingScenes2022} have demonstrated +\autocite{mildenhallNeRFRepresentingScenes2022} have demonstrated exceptional rendering quality but suffer from computational inefficiencies due to per-pixel ray marching in volume rendering, making real-time applications challenging. The recent development of \textbf{3D Gaussian Splatting} -\cite{kerbl3DGaussianSplatting2023} offers a promising alternative +\autocite{kerbl3DGaussianSplatting2023} offers a promising alternative by employing a rasterization-based rendering pipeline. In this method, scenes are represented using a set of 3D Gaussians, which can be efficiently projected onto the image plane and rasterized to produce @@ -426,7 +411,7 @@ \section{Method}\label{method} \subsection{Scene Representation}\label{scene-representation} Building upon the Gaussian splatting method -\cite{kerbl3DGaussianSplatting2023}, we adapt the scene +\autocite{kerbl3DGaussianSplatting2023}, we adapt the scene representation to focus on the differentiable depth rendering process, which is crucial for our localization task. Our approach utilizes the efficiency and quality of Gaussian splatting while tailoring it @@ -453,7 +438,7 @@ \subsection{Scene Representation}\label{scene-representation} \textbf{Projecting 3D to 2D.} For the projection of 3D Gaussians onto the 2D image plane, we follow the approach described by -\cite{kerbl3DGaussianSplatting2023}. The 3D mean +\autocite{kerbl3DGaussianSplatting2023}. The 3D mean \(\boldsymbol{\mu}_i\) is first transformed into the camera coordinate frame using the world-to-camera transformation \(\mathbf{T}_{wc} \in SE(3)\). Then, it is projected using the @@ -475,7 +460,7 @@ \subsection{Scene Representation}\label{scene-representation} where \(\mathbf{R}_{wc}\) represents the rotation component of \(\mathbf{T}_{wc}\), and \(\mathbf{J}\) is the Jacobian of the projection function, accounting for the affine transformation from 3D to -2D as described by \cite{zwickerEWASplatting2002}. +2D as described by \autocite{zwickerEWASplatting2002}. \subsection{Depth Rendering}\label{depth-rendering} @@ -490,7 +475,7 @@ \subsection{Depth Rendering}\label{depth-rendering} depth value of the \(n\)-th Gaussian, corresponding to the z-coordinate of its mean in the camera coordinate system. The depth at pixel \(\mathbf{p}\), denoted \(D(\mathbf{p})\), is computed as -\cite{kerbl3DGaussianSplatting2023}: +\autocite{kerbl3DGaussianSplatting2023}: \[D(\mathbf{p}) = \sum_{n \leq N} d_n \cdot \alpha_n \cdot T_n,\] @@ -548,7 +533,7 @@ \subsection{Localization as Image representation and the query depth image. \textbf{Rotating with -Quaternions.}\cite{kuipersQuaternionsRotationSequences1999} We +Quaternions.}\autocite{kuipersQuaternionsRotationSequences1999} We parameterize the camera pose using a quaternion \(\mathbf{q}_{cw}\) for rotation and a vector \(\mathbf{t}_{cw}\) for translation. This choice of parameterization is particularly advantageous in our differential @@ -584,7 +569,7 @@ \subsection{Localization as Image \] Here, \(\nabla D\) represents the gradient of the depth image, computed -using the Sobel operator \cite{kanopoulosDesignImageEdge1988}, and +using the Sobel operator \autocite{kanopoulosDesignImageEdge1988}, and \(\mathcal{M}\) is the mask of valid pixels determined by the rendered alpha mask. @@ -658,14 +643,17 @@ \subsection{Pipeline}\label{pipeline} objective function. \textbf{Optimization.} We employ the -Adam\cite{kingmaAdamMethodStochastic2014} optimizer for optimizing +Adam\autocite{kingmaAdamMethodStochastic2014} optimizer for optimizing both the quaternion and translation parameters, using the distinct learning rates and weight decay values as previously described. The optimization process greatly benefits from the real-time rendering capabilities of 3D Gaussian splatting. Since rendering is extremely fast, each iteration of the optimizer is limited mainly by the rendering speed, allowing for rapid convergence of our pose estimation algorithm -and making it suitable for real-time applications. +and making it suitable for real-time applications. Our optimization +approach consistently achieves sub-millimeter accuracy (average ATE RMSE +of \textbf{0.01587 cm}) on synthetic datasets, while maintaining robust +performance in real-world scenarios. \textbf{Convergence.} To determine convergence, we implement an early stopping mechanism based on the stabilization of the total loss. Our @@ -682,12 +670,13 @@ \subsection{Pipeline}\label{pipeline} \section{Evaluation}\label{evaluation} -We conducted extensive experiments to evaluate the performance of our -proposed method, \textbf{GSplatLoc}, in comparison with state-of-the-art -SLAM systems that utilize advanced scene representations. The evaluation -focuses on assessing the accuracy of camera pose estimation in -challenging indoor environments, emphasizing both the translational and -rotational components of the estimated poses. +We conducted a comprehensive evaluation spanning both synthetic and +real-world environments, with pose estimation errors ranging from as low +as \textbf{0.01587 cm} in controlled settings to competitive performance +(\textbf{0.80982 cm}) in challenging real-world scenarios. Our +evaluation framework encompasses multiple aspects of localization +performance, from implementation details to dataset selection and +baseline comparisons. \subsection{Experimental Setup}\label{experimental-setup} @@ -702,15 +691,15 @@ \subsection{Experimental Setup}\label{experimental-setup} \textbf{Datasets.} We evaluated our method on two widely recognized datasets for SLAM benchmarking: the \textbf{Replica} dataset -\cite{straubReplicaDatasetDigital2019} and the \textbf{TUM RGB-D} -dataset \cite{sturmBenchmarkEvaluationRGBD2012}. The Replica dataset +\autocite{straubReplicaDatasetDigital2019} and the \textbf{TUM RGB-D} +dataset \autocite{sturmBenchmarkEvaluationRGBD2012}. The Replica dataset provides high-fidelity synthetic indoor environments, ideal for controlled evaluations of localization algorithms. We utilized data -collected by Sucar et al. \cite{sucarImapImplicitMapping2021}, which +collected by Sucar et al. \autocite{sucarImapImplicitMapping2021}, which includes trajectories from an RGB-D sensor with ground-truth poses. The -TUM RGB-D dataset offers real-world sequences captured in various indoor -settings, providing a diverse range of scenarios to test the robustness -of our method. +TUM RGB-D dataset\autocite{sturmBenchmarkEvaluationRGBD2012} offers +real-world sequences captured in various indoor settings, providing a +diverse range of scenarios to test the robustness of our method. \textbf{Metrics.} Localization accuracy was assessed using two standard metrics: the \textbf{Absolute Trajectory Error (ATE RMSE)}, measured in @@ -722,13 +711,13 @@ \subsection{Experimental Setup}\label{experimental-setup} \textbf{Baselines.}~To provide a comprehensive comparison, we evaluated our method against several state-of-the-art SLAM systems that leverage advanced scene representations. Specifically, we compared against -RTG-SLAM (ICP) \cite{pengRTGSLAMRealtime3D2024}, which utilizes +RTG-SLAM(ICP)\autocite{pengRTGSLAMRealtime3D2024}, which utilizes Iterative Closest Point (ICP) for pose estimation within a 3D Gaussian -splatting framework. We also included GS-ICP-SLAM (GICP) -\cite{haRGBDGSICPSLAM2024}, which employs Generalized ICP for +splatting framework. We also included GS-ICP-SLAM(GICP) +\autocite{haRGBDGSICPSLAM2024}, which employs Generalized ICP for alignment in a Gaussian-based representation. Additionally, we considered Gaussian-SLAM -\cite{yugayGaussianSLAMPhotorealisticDense2024}, evaluating both its +\autocite{yugayGaussianSLAMPhotorealisticDense2024}, evaluating both its PLANE ICP and HYBRID variants, which adapt traditional RGB-D odometry methods by incorporating plane-based ICP and a hybrid approach combining photometric and geometric information. These baselines were selected @@ -738,96 +727,88 @@ \subsection{Experimental Setup}\label{experimental-setup} \subsection{Localization Evaluation}\label{localization-evaluation} -We first evaluated our method on the Replica dataset, which provides a -controlled environment to assess the accuracy of pose estimation -algorithms. +We conducted comprehensive experiments on both synthetic and real-world +datasets to evaluate the performance of GSplatLoc against +state-of-the-art methods utilizing advanced scene representations. \begin{table}[htbp] \renewcommand{\thetable}{\textbf{\arabic{table}}} \renewcommand{\tablename}{\textbf{Table}} \centering \caption{\textbf{Replica\cite{straubReplicaDatasetDigital2019} (ATE RMSE ↓[cm]).}} -\label{table:replica_ATE} +\label{table:_textbf_replica_cite} \begin{adjustbox}{max width=\columnwidth,max height=!,center} \begin{tabular}{lccccccccc} \toprule \textbf{Methods} & \textbf{Avg.} & \textbf{R0} & \textbf{R1} & \textbf{R2} & \textbf{Of0} & \textbf{Of1} & \textbf{Of2} & \textbf{Of3} & \textbf{Of4}\\ \midrule -RTG-SLAM(ICP)\cite{pengRTGSLAMRealtime3D2024} & \cellcolor{yellow!30}0.471 & \cellcolor{yellow!30}0.429 & \cellcolor{yellow!30}0.690 & \cellcolor{yellow!30}0.544 & \cellcolor{yellow!30}0.640 & \cellcolor{yellow!30}0.336 & \cellcolor{yellow!30}0.434 & \cellcolor{yellow!30}0.281 & \cellcolor{yellow!30}0.419\\ -GS-ICP-SLAM(GICP)\cite{haRGBDGSICPSLAM2024} & \cellcolor{lime!50}0.593 & \cellcolor{lime!50}0.465 & \cellcolor{lime!50}0.772 & \cellcolor{lime!50}0.723 & \cellcolor{lime!50}0.681 & \cellcolor{lime!50}0.522 & \cellcolor{lime!50}0.582 & \cellcolor{lime!50}0.438 & \cellcolor{lime!50}0.558\\ -Gaussian-SLAM(PLANE ICP)\cite{yugayGaussianSLAMPhotorealisticDense2024} & 0.633 & 0.476 & 0.812 & 0.781 & 0.709 & 0.541 & 0.667 & 0.449 & 0.625\\ -Gaussian-SLAM(HYBRID)\cite{yugayGaussianSLAMPhotorealisticDense2024} & 0.631 & 0.476 & 0.812 & 0.781 & 0.709 & 0.537 & 0.662 & 0.446 & 0.624\\ +RTG-SLAM(ICP)\cite{pengRTGSLAMRealtime3D2024} & 1.102 & 1.286 & 0.935 & \cellcolor{yellow!30}1.117 & 0.983 & 0.626 & 1.194 & \cellcolor{yellow!30}1.334 & 1.340\\ +GS-ICP-SLAM(GICP)\cite{haRGBDGSICPSLAM2024} & \cellcolor{yellow!30}1.084 & 1.250 & \cellcolor{yellow!30}0.828 & 1.183 & \cellcolor{lime!50}0.924 & \cellcolor{lime!50}0.591 & \cellcolor{lime!50}1.175 & 1.438 & \cellcolor{yellow!30}1.284\\ +Gaussian-SLAM(PLANE ICP)\cite{yugayGaussianSLAMPhotorealisticDense2024} & \cellcolor{lime!50}1.086 & \cellcolor{yellow!30}1.246 & 0.855 & 1.186 & \cellcolor{yellow!30}0.922 & \cellcolor{yellow!30}0.590 & \cellcolor{yellow!30}1.162 & \cellcolor{lime!50}1.426 & 1.304\\ +Gaussian-SLAM(HYBRID)\cite{yugayGaussianSLAMPhotorealisticDense2024} & 1.096 & \cellcolor{lime!50}1.248 & \cellcolor{lime!50}0.831 & \cellcolor{lime!50}1.183 & 0.926 & 0.595 & 1.201 & 1.499 & \cellcolor{lime!50}1.289\\ \midrule -\textbf{Ours} & \cellcolor{green!30}\textbf{0.009} & \cellcolor{green!30}\textbf{0.007} & \cellcolor{green!30}\textbf{0.008} & \cellcolor{green!30}\textbf{0.010} & \cellcolor{green!30}\textbf{0.009} & \cellcolor{green!30}\textbf{0.009} & \cellcolor{green!30}\textbf{0.011} & \cellcolor{green!30}\textbf{0.009} & \cellcolor{green!30}\textbf{0.011}\\ +\textbf{Ours} & \cellcolor{green!30}\textbf{0.016} & \cellcolor{green!30}\textbf{0.015} & \cellcolor{green!30}\textbf{0.013} & \cellcolor{green!30}\textbf{0.021} & \cellcolor{green!30}\textbf{0.011} & \cellcolor{green!30}\textbf{0.009} & \cellcolor{green!30}\textbf{0.018} & \cellcolor{green!30}\textbf{0.020} & \cellcolor{green!30}\textbf{0.019}\\ \bottomrule \end{tabular} \end{adjustbox} \end{table} -\textbf{Table 1.} presents the ATE RMSE results in centimeters for -various methods across different sequences in the Replica dataset. Our -method significantly outperforms the baselines, achieving an average ATE -RMSE of \textbf{0.00925 cm}, which is an order of magnitude better than -the closest competitor. This substantial improvement demonstrates the -effectiveness of our approach in accurately estimating the camera's -position. The low translational errors indicate that our method can -precisely align the observed depth images with the rendered depth from -the 3D Gaussian scene. +\textbf{Table 1.} presents the Absolute Trajectory Error (ATE RMSE) +results on the Replica dataset. Our method achieves remarkable +performance with an average ATE RMSE of \textbf{0.01587 cm}, +significantly outperforming existing approaches by nearly two orders of +magnitude. The closest competitor, RTG-SLAM(ICP) +\autocite{pengRTGSLAMRealtime3D2024}, achieves an average error of +1.10186 cm. This substantial improvement is consistent across all +sequences, with particularly notable performance in challenging scenes +like Of1 (0.00937 cm) and R1 (0.01272 cm). \begin{table}[htbp] \renewcommand{\thetable}{\textbf{\arabic{table}}} \renewcommand{\tablename}{\textbf{Table}} \centering \caption{\textbf{Replica\cite{straubReplicaDatasetDigital2019} (AAE RMSE ↓[°]).}} -\label{table:replica_AAE} +\label{table:_textbf_replica_cite} \begin{adjustbox}{max width=\columnwidth,max height=!,center} \begin{tabular}{lccccccccc} \toprule \textbf{Methods} & \textbf{Avg.} & \textbf{R0} & \textbf{R1} & \textbf{R2} & \textbf{Of0} & \textbf{Of1} & \textbf{Of2} & \textbf{Of3} & \textbf{Of4}\\ \midrule -RTG-SLAM(ICP)\cite{pengRTGSLAMRealtime3D2024} & \cellcolor{green!30}\textbf{0.576} & \cellcolor{green!30}\textbf{0.720} & \cellcolor{green!30}\textbf{0.826} & \cellcolor{yellow!30}0.744 & \cellcolor{green!30}\textbf{0.054} & \cellcolor{green!30}\textbf{0.537} & \cellcolor{yellow!30}0.360 & \cellcolor{yellow!30}0.330 & \cellcolor{yellow!30}0.430\\ -GS-ICP-SLAM(GICP)\cite{haRGBDGSICPSLAM2024} & \cellcolor{lime!50}1.279 & \cellcolor{lime!50}1.659 & 1.951 & 1.607 & \cellcolor{lime!50}0.281 & \cellcolor{yellow!30}0.895 & 2.580 & 1.110 & 2.940\\ -Gaussian-SLAM(PLANE ICP)\cite{yugayGaussianSLAMPhotorealisticDense2024} & 1.287 & 1.834 & \cellcolor{lime!50}1.880 & \cellcolor{lime!50}1.398 & 0.305 & 1.019 & 1.060 & 1.100 & 1.130\\ -Gaussian-SLAM(HYBRID)\cite{yugayGaussianSLAMPhotorealisticDense2024} & 1.955 & 2.265 & 3.493 & 2.783 & 0.287 & \cellcolor{lime!50}0.945 & \cellcolor{lime!50}0.580 & \cellcolor{lime!50}0.720 & \cellcolor{lime!50}0.630\\ +RTG-SLAM(ICP)\cite{pengRTGSLAMRealtime3D2024} & \cellcolor{yellow!30}0.471 & \cellcolor{yellow!30}0.429 & \cellcolor{yellow!30}0.690 & \cellcolor{yellow!30}0.544 & \cellcolor{yellow!30}0.640 & \cellcolor{yellow!30}0.336 & \cellcolor{yellow!30}0.434 & \cellcolor{yellow!30}0.281 & \cellcolor{yellow!30}0.419\\ +GS-ICP-SLAM(GICP)\cite{haRGBDGSICPSLAM2024} & 0.631 & 0.476 & 0.812 & 0.781 & 0.709 & 0.537 & 0.662 & 0.446 & 0.624\\ +Gaussian-SLAM(PLANE ICP)\cite{yugayGaussianSLAMPhotorealisticDense2024} & \cellcolor{lime!50}0.593 & \cellcolor{lime!50}0.465 & \cellcolor{lime!50}0.772 & \cellcolor{lime!50}0.723 & \cellcolor{lime!50}0.681 & \cellcolor{lime!50}0.522 & \cellcolor{lime!50}0.582 & \cellcolor{lime!50}0.438 & \cellcolor{lime!50}0.558\\ +Gaussian-SLAM(HYBRID)\cite{yugayGaussianSLAMPhotorealisticDense2024} & 0.633 & 0.476 & 0.812 & 0.781 & 0.709 & 0.541 & 0.667 & 0.449 & 0.625\\ \midrule -\textbf{Ours} & \cellcolor{yellow!30}0.810 & \cellcolor{yellow!30}0.931 & \cellcolor{yellow!30}1.006 & \cellcolor{green!30}\textbf{0.666} & \cellcolor{yellow!30}0.248 & 1.197 & \cellcolor{green!30}\textbf{0.011} & \cellcolor{green!30}\textbf{0.009} & \cellcolor{green!30}\textbf{0.011}\\ +\textbf{Ours} & \cellcolor{green!30}\textbf{0.009} & \cellcolor{green!30}\textbf{0.007} & \cellcolor{green!30}\textbf{0.008} & \cellcolor{green!30}\textbf{0.010} & \cellcolor{green!30}\textbf{0.009} & \cellcolor{green!30}\textbf{0.009} & \cellcolor{green!30}\textbf{0.011} & \cellcolor{green!30}\textbf{0.009} & \cellcolor{green!30}\textbf{0.011}\\ \bottomrule \end{tabular} \end{adjustbox} \end{table} -\textbf{Table 2.} presents the Absolute Angular Error (AAE) RMSE in -degrees for various methods on the Replica dataset. Our method achieves -a competitive average AAE RMSE of \textbf{0.80982°}, indicating superior -rotational accuracy in most sequences. In sequences with significant -rotational movements, such as Of2, Of3, and Of4, our approach -consistently outperforms the baselines. For instance, in sequence Of3, -our method achieves an AAE RMSE of \textbf{0.00930°}, compared to -\textbf{0.33000°} by RTG-SLAM and higher errors by other methods. This -exceptional performance can be attributed to the effective utilization -of the differentiable rendering pipeline and the optimization strategy -that precisely aligns the depth gradients between the rendered and -observed images. - -To evaluate the robustness of our method in real-world scenarios, we -conducted experiments on the TUM RGB-D dataset, which presents -challenges such as sensor noise and dynamic environments. +\textbf{Table 2.} GSplatLoc achieves an average AAE RMSE of +\textbf{0.00925°}. This represents a significant improvement over +traditional ICP-based methods, with +RTG-SLAM\autocite{pengRTGSLAMRealtime3D2024} and +GS-ICP-SLAM\autocite{haRGBDGSICPSLAM2024} showing average errors of +0.47141° and 0.63100° respectively. The performance advantage is +particularly evident in sequences with complex rotational movements, +such as Of2 and Of4, where our method maintains sub-0.01° accuracy. \begin{table}[htbp] \renewcommand{\thetable}{\textbf{\arabic{table}}} \renewcommand{\tablename}{\textbf{Table}} \centering \caption{\textbf{TUM\cite{sturmBenchmarkEvaluationRGBD2012} (ATE RMSE ↓[cm]).}} -\label{table:tum_ATE} +\label{table:_textbf_tum_cite_stu} \begin{adjustbox}{max width=\columnwidth,max height=!,center} \begin{tabular}{lcccccc} \toprule \textbf{Methods} & \textbf{Avg.} & \textbf{fr1/desk} & \textbf{fr1/desk2} & \textbf{fr1/room} & \textbf{fr2/xyz} & \textbf{fr3/off.}\\ \midrule RTG-SLAM(ICP)\cite{pengRTGSLAMRealtime3D2024} & \cellcolor{green!30}\textbf{0.576} & \cellcolor{green!30}\textbf{0.720} & \cellcolor{green!30}\textbf{0.826} & \cellcolor{yellow!30}0.744 & \cellcolor{green!30}\textbf{0.054} & \cellcolor{green!30}\textbf{0.537}\\ -GS-ICP-SLAM(GICP)\cite{haRGBDGSICPSLAM2024} & \cellcolor{lime!50}1.279 & \cellcolor{lime!50}1.659 & 1.951 & 1.607 & \cellcolor{lime!50}0.281 & \cellcolor{yellow!30}0.895\\ -Gaussian-SLAM(PLANE ICP)\cite{yugayGaussianSLAMPhotorealisticDense2024} & 1.287 & 1.834 & \cellcolor{lime!50}1.880 & \cellcolor{lime!50}1.398 & 0.305 & 1.019\\ -Gaussian-SLAM(HYBRID)\cite{yugayGaussianSLAMPhotorealisticDense2024} & 1.955 & 2.265 & 3.493 & 2.783 & 0.287 & \cellcolor{lime!50}0.945\\ +GS-ICP-SLAM(GICP)\cite{haRGBDGSICPSLAM2024} & 1.955 & 2.265 & 3.493 & 2.783 & 0.287 & \cellcolor{lime!50}0.945\\ +Gaussian-SLAM(PLANE ICP)\cite{yugayGaussianSLAMPhotorealisticDense2024} & \cellcolor{lime!50}1.279 & \cellcolor{lime!50}1.659 & 1.951 & 1.607 & \cellcolor{lime!50}0.281 & \cellcolor{yellow!30}0.895\\ +Gaussian-SLAM(HYBRID)\cite{yugayGaussianSLAMPhotorealisticDense2024} & 1.287 & 1.834 & \cellcolor{lime!50}1.880 & \cellcolor{lime!50}1.398 & 0.305 & 1.019\\ \midrule \textbf{Ours} & \cellcolor{yellow!30}0.810 & \cellcolor{yellow!30}0.931 & \cellcolor{yellow!30}1.006 & \cellcolor{green!30}\textbf{0.666} & \cellcolor{yellow!30}0.248 & 1.197\\ \bottomrule @@ -836,34 +817,45 @@ \subsection{Localization Evaluation}\label{localization-evaluation} \end{table} \textbf{Table 3.} presents the ATE RMSE in centimeters for various -methods on the TUM-RGBD dataset -\cite{sturmBenchmarkEvaluationRGBD2012}. Our method achieves -competitive results with an average ATE RMSE of \textbf{8.0982 cm}, -outperforming GS-ICP-SLAM\cite{haRGBDGSICPSLAM2024} and -Gaussian-SLAM\cite{yugayGaussianSLAMPhotorealisticDense2024} in most -sequences. While RTG-SLAM\cite{pengRTGSLAMRealtime3D2024} shows +methods on the TUM-RGBD dataset . Our method achieves competitive +results with an average ATE RMSE of \textbf{8.0982 cm}, outperforming +GS-ICP-SLAM\autocite{haRGBDGSICPSLAM2024} and +Gaussian-SLAM\autocite{yugayGaussianSLAMPhotorealisticDense2024} in most +sequences. While RTG-SLAM\autocite{pengRTGSLAMRealtime3D2024} shows lower errors in some sequences, our method consistently provides accurate pose estimates across different environments. The increased error compared to the Replica dataset is expected due to the real-world -challenges present in the TUM RGB-D dataset, such as sensor noise and -environmental variability. Despite these challenges, our method +challenges present in the TUM RGB-D +dataset\autocite{sturmBenchmarkEvaluationRGBD2012}, such as sensor noise +and environmental variability. Despite these challenges, our method demonstrates robustness and maintains reasonable localization accuracy. +\textbf{Tables 3.} presents results on the more challenging TUM RGB-D +dataset\autocite{sturmBenchmarkEvaluationRGBD2012}, which introduces +real-world complexities such as sensor noise and dynamic environments. +In terms of translational accuracy, GSplatLoc achieves competitive +performance with an average ATE RMSE of \textbf{0.80982 cm}. While +RTG-SLAM\autocite{pengRTGSLAMRealtime3D2024} shows slightly better +average performance (0.57636 cm), our method consistently outperforms +both GS-ICP-SLAM\autocite{haRGBDGSICPSLAM2024} (1.95454 cm) and +Gaussian-SLAM\autocite{yugayGaussianSLAMPhotorealisticDense2024} +variants (1.27873 cm and 1.28716 cm) across most sequences. + \begin{table}[htbp] \renewcommand{\thetable}{\textbf{\arabic{table}}} \renewcommand{\tablename}{\textbf{Table}} \centering \caption{\textbf{TUM\cite{sturmBenchmarkEvaluationRGBD2012} (AAE RMSE ↓[°]).}} -\label{table:tum_AAE} +\label{table:_textbf_tum_cite_stu} \begin{adjustbox}{max width=\columnwidth,max height=!,center} \begin{tabular}{lcccccc} \toprule \textbf{Methods} & \textbf{Avg.} & \textbf{fr1/desk} & \textbf{fr1/desk2} & \textbf{fr1/room} & \textbf{fr2/xyz} & \textbf{fr3/off.}\\ \midrule RTG-SLAM(ICP)\cite{pengRTGSLAMRealtime3D2024} & \cellcolor{green!30}\textbf{0.916} & \cellcolor{yellow!30}1.181 & \cellcolor{yellow!30}1.557 & \cellcolor{yellow!30}1.355 & \cellcolor{yellow!30}0.138 & \cellcolor{green!30}\textbf{0.347}\\ -GS-ICP-SLAM(GICP)\cite{haRGBDGSICPSLAM2024} & \cellcolor{yellow!30}0.959 & \cellcolor{lime!50}1.288 & \cellcolor{lime!50}1.618 & \cellcolor{lime!50}1.363 & \cellcolor{lime!50}0.147 & \cellcolor{lime!50}0.381\\ -Gaussian-SLAM(PLANE ICP)\cite{yugayGaussianSLAMPhotorealisticDense2024} & 1.090 & 1.388 & 1.791 & 1.564 & 0.182 & 0.525\\ -Gaussian-SLAM(HYBRID)\cite{yugayGaussianSLAMPhotorealisticDense2024} & 1.117 & 1.426 & 2.098 & 1.594 & \cellcolor{green!30}\textbf{0.114} & \cellcolor{yellow!30}0.355\\ +GS-ICP-SLAM(GICP)\cite{haRGBDGSICPSLAM2024} & 1.117 & 1.426 & 2.098 & 1.594 & \cellcolor{green!30}\textbf{0.114} & \cellcolor{yellow!30}0.355\\ +Gaussian-SLAM(PLANE ICP)\cite{yugayGaussianSLAMPhotorealisticDense2024} & \cellcolor{yellow!30}0.959 & \cellcolor{lime!50}1.288 & \cellcolor{lime!50}1.618 & \cellcolor{lime!50}1.363 & \cellcolor{lime!50}0.147 & \cellcolor{lime!50}0.381\\ +Gaussian-SLAM(HYBRID)\cite{yugayGaussianSLAMPhotorealisticDense2024} & 1.090 & 1.388 & 1.791 & 1.564 & 0.182 & 0.525\\ \midrule \textbf{Ours} & \cellcolor{lime!50}0.979 & \cellcolor{green!30}\textbf{1.126} & \cellcolor{green!30}\textbf{1.265} & \cellcolor{green!30}\textbf{0.907} & 0.789 & 0.808\\ \bottomrule @@ -871,16 +863,22 @@ \subsection{Localization Evaluation}\label{localization-evaluation} \end{adjustbox} \end{table} -\textbf{Table 4.} presents the AAE RMSE results in degrees for the TUM -RGB-D dataset. Our method achieves an average AAE RMSE of -\textbf{0.97928°}, which is competitive with the other methods. In -sequences such as fr1/room, our method demonstrates superior rotational -accuracy with an AAE RMSE of \textbf{0.90722°}, compared to higher -errors by the baselines. The slightly higher rotational errors in the -TUM RGB-D dataset, compared to the Replica dataset, can be attributed to -the complexities of real-world data, including sensor inaccuracies and -dynamic elements in the environment. Nonetheless, our method maintains -reliable performance across various sequences. +\textbf{Table 4.} The rotational accuracy results on TUM RGB-D +dataset\autocite{sturmBenchmarkEvaluationRGBD2012} demonstrate the +robustness of our approach in real-world scenarios. GSplatLoc maintains +stable performance with an average AAE RMSE of \textbf{0.97928°}, +comparable to RTG-SLAM's 0.91561°. Notably, our method shows superior +performance in challenging sequences like fr1/room (0.90722°) compared +to competing methods, which exhibit errors ranging from 1.35470° to +1.56370°. + +The performance gap between synthetic and real-world results highlights +the impact of sensor noise and environmental complexity on localization +accuracy. While the near-perfect accuracy achieved on the Replica +dataset demonstrates the theoretical capabilities of our approach, the +competitive performance on TUM RGB-D +dataset\autocite{sturmBenchmarkEvaluationRGBD2012} validates its +practical applicability in real-world scenarios. \subsection{Discussion}\label{discussion} @@ -906,11 +904,12 @@ \subsection{Discussion}\label{discussion} optimization process. While our method shows excellent performance on the Replica dataset, the -increased errors on the TUM RGB-D dataset highlight areas for potential -improvement. Real-world datasets introduce challenges such as sensor -noise, dynamic objects, and incomplete depth data due to occlusions. -Addressing these challenges in future work could further enhance the -robustness of our method. +increased errors on the TUM RGB-D +dataset\autocite{sturmBenchmarkEvaluationRGBD2012} highlight areas for +potential improvement. Real-world datasets introduce challenges such as +sensor noise, dynamic objects, and incomplete depth data due to +occlusions. Addressing these challenges in future work could further +enhance the robustness of our method. \subsection{Limitations}\label{limitations} @@ -936,14 +935,16 @@ \section{Conclusion}\label{conclusion} alignment between rendered depth maps from a pre-existing 3D Gaussian scene and observed depth images. -Extensive experiments on the Replica and TUM RGB-D datasets demonstrate -that GSplatLoc significantly outperforms state-of-the-art SLAM systems -in terms of both translational and rotational accuracy. On the Replica +Extensive experiments on the Replica and TUM RGB-D +dataset\autocite{sturmBenchmarkEvaluationRGBD2012} demonstrate that +GSplatLoc significantly outperforms state-of-the-art SLAM systems in +terms of both translational and rotational accuracy. On the Replica dataset, our method achieves an average Absolute Trajectory Error (ATE -RMSE) of 0.00925\,cm, surpassing existing approaches by an order of -magnitude. The method also maintains competitive performance on the TUM -RGB-D dataset, exhibiting robustness in real-world scenarios despite -challenges such as sensor noise and dynamic elements. +RMSE) of \textbf{0.01587 cm}, surpassing existing approaches by an order +of magnitude. The method also maintains competitive performance on the +TUM RGB-D dataset\autocite{sturmBenchmarkEvaluationRGBD2012}, exhibiting +robustness in real-world scenarios despite challenges such as sensor +noise and dynamic elements. The superior performance of GSplatLoc can be attributed to several key factors. The utilization of a fully differentiable depth rendering @@ -973,5 +974,3 @@ \section{Conclusion}\label{conclusion} \printbibliography \end{document} - -