Skip to content

Latest commit

 

History

History
8 lines (8 loc) · 751 Bytes

README.md

File metadata and controls

8 lines (8 loc) · 751 Bytes
Raw
  1. clone yolo repository: http://github.com/alexeyab/darknet
  2. replace existing data dir with data dir from this repo
  3. run the convert_pig_to_yolo.py script to convert from our existing annotation format to yolo's format
  4. run the create_train_test_split.py to split up into train/test sets
  5. modify Makefile and set GPU=1, CUDNN=1, LIBSO=1, and uncomment ARCH= line corresponding to whatever GPU you have
  6. run make in the current directory
  7. to train, download existing yolo weights, and run ./darknet detector train data/obj.data cfg/yolo-obj.cfg yolov4.conv.137.
  8. to annotate, run python3 annotate.py --input INPUT_FILE