Starting from scratch
python pipeline/train_boulder_segmentation.py -f
Loading a specific checkpoint version
python pipeline/train_boulder_segmentation.py -c /path/to/.pth
If no argument is given the training continues from "workdir/latest.pth"
python pipeline/test_boulder_segementation.py --show-dir /data/results
--eval mIou -p /path/to/.pth
all arguments optional. If no checkpoint file (-p) is given "workdir/latest.pth" is used.
python pipeline/test_boulder_segementation.py --show-dir /data/results
--eval mIou -p /path/to/.pth