README-Splits.md

---

Home     |     Data format     |     Downloads     |     Data loader     |     Challenges     |     Leaderboards

---

Dataset Splits and Baselines

The official splits of the datasets for deep learning models are provided in the paper [1]: The paper provides extended experiments on the baseline methods using the official dataset splits.

Fig. 1: Cross-validation experiment schedule for CholecT50. For CholecT45, remove the last video in each fold. The number in each box represents the video ID. .

Video Overlap

• As the dataset is from CAMMA research group, University of Strasbourg, France, there are possible video overlaps in other cholecystectomy datasets such as Cholec80, Cholec120, M2CAI16, etc.
• The video IDs (e.g. 1, 2, 5, 80, etc.) are consistent across these datasets. The prefix "VID" in the video filenames (e.g. VID01, VID02, VID80, etc.) are sometimes written as "Video" in other datasets (e.g. Video01, Video80, etc. )
• Researchers are advised to take into consideration the overlapping videos when pre-training their models on other cholecystectomy datasets.

---

References

• [1] C.I. Nwoye, N. Padoy. Data Splits and Metrics for Benchmarking Methods on Surgical Action Triplet Datasets. arXiv PrePrint arXiv:2204.05235. 2022.

  @article{nwoye2022data,
    title={Data Splits and Metrics for Benchmarking Methods on Surgical Action Triplet Datasets},
    author={Nwoye, Chinedu Innocent and Padoy, Nicolas},
    journal={arXiv preprint arXiv:2204.05235},
    year={2022}
  }