Designed to create training data for: https://github.com/AlexeyAB/darknet.git
-
Put it python script in a folder. Let's name it "NDDS_data". You can of course name it as you want :)
-
Add yolov3.cfg to a folder (you can find it in darknet/cfg folder)
-
Rename "yolov3.cfg" to "yolo-obj.cfg"
-
Put train images to /NDDS_data/data_train folder
-
Put validation images to /NDDS_data/data_valid folder
-
Put "darknet" compiled file in NDDS_data folder
-
Put "data/labels" folders from darknet to NDDS_data folder
-
Type in terminal: python create_dataset.py
-
Wait till it's done. There might be some warnings during this process. Don't worry :D
-
Now you can just run_training.sh script to start training your model
-
To modify parameters of training just edit yolo-obj.cfg file
-
To run training you can use run_training.sh script. Just put it in NDDS_data folder and run it!
-
Wait till you teach your network
-
Enjoy!