Annotation system for labelling bounding boxes using openCV
To install, just requires numpy and cv2:
pip install numpy opencv-python
Edit config.py to point at the correct directory for json files and images, and specify wherever you want the labels to be saved.
Then just run annotate_json.py. It will ask you which set you want to annotate, and open an interactive cv2 window.
You will be presented with an image, and the relevant question-answer pair will be shown in the terminal.
You can draw bounding boxes by clicking and dragging, and delete bounding boxes by right-clicking on them or pressing d.
Press n to go to the next image, p for the previous image, or q to finish and exit the program.
cv2 does not behave very well if you close the image window manually or attempt to KeyboardInterrupt out, so using q to close the window is recommended.
Annotations are saved as .npy arrays in the specified label directory, named after the questionIds defined in the json file.
Please let me know if you encounter any problems.