This codebase is developed and maintained by Saket Mohanty, This code base uses mlcv framework, which is build by Saket and will be make available through pypi.
- Use the environment.yml to create conda environment
- First Train the model.
source run.sh - Now to test the model you just need to
python infer_ckpt.py --image_dir <PATH_TO_THE_DIR_CONTAINING_VIDEO> --output_dir <PATH_TO_THE_DIR_WHERE_OUTPUT_VIDEO_WILL_BE_STORED> --restore_from <PATH_TO_CKPT>
The trained model for this model can be found here.
IOU on single image for the model provided is 71%.