BoundingBoxTool with deep learning support to classify objects on images, at the first version will support darknet models. This work is inspired by LabelImg: Git code (2015). https://github.com/tzutalin/labelImg
Software dependencies: *OpenCV *Gtk3.0 *cairo *darknet
The execution is: python3 dbbt.py
- execute the program with:
python3 dbbt.py
- press the select Path button on the top right to select the model Path:
- Select the folder that contain the darknet model files (.cfg, .weights, .data, .names) including libdarknet.so, be sure of include the .names folder correctly on the .data file, the .so file is generated when compile darknet, for this work is used the darknet version of AlexeyAB AlexeyAB Darknet, you must move the .so generated of the darknet compilation to the net folder, remember to put in this same folder the .cfg, .weights, .data and .names files.
- If all is working fine so you must se the files loaded, on the terminal must be loaded the darknet model without errors.
- Press the button open folder to open images folder path
- Press open to open the folder
- All is ready to start to work!!.
- [] easy selection of working labels when creating a bounding box, ie. improve popover of classes suggestion.
- [] establish a method to save and record when an image has been processed.
- [] create log file to remember selected folder previously and doesn't set folder paths every time the program is opened.
- [] establish a better user experience and keep all menu buttons working
- [] with previous steps covered, this software can be considered in an stable release.
- [] VOC-PASCAL Data format: create data annotations in this format
- [] accept others deep learning networks architectures and use with TensorFlow, Caffe, ONNX, etc, etc... im opened to suggestions.
if you want to contribute you can mail me to carlostrujillotmx@gmail.com