edwinlimlx / darknet-capture Goto Github PK

YOLO V2

1. Create folder for darknet images and label

mkdir captures captures/edwin captures/edwin/final

1. Change the name for class in src/image.c line 226

2. Change the path in cfg/yolo-face.data

3. Change the name in data/face.names

4. Compile the darknet executable

make clean && make

5. Start the capture Try with -thresh .7 or more for better quality training. Using yolo-face.cfg and yolo-face_final.weights from quanhua92's weights https://mega.nz/#F!GRV1XKbJ!v8BCsFO8iJVNppiGXY4qMw

clear && ./darknet detector demo cfg/yolo-face.data cfg/yolo-face.cfg weights/yolo-face_final.weights -thresh .6

6. Shorthand for recompiling and start capturing

clear && make clean && make && ./darknet detector demo cfg/yolo-face.data weights/yolo-face.cfg weights/yolo-face_final.weights -thresh .6

./darknet yolo train cfg/yolo-face.cfg weights/yolo-face_final.weights
./darknet detector train cfg/yolo-face.data cfg/yolo-face.cfg weights/darknet19_448.conv.23

./darknet detector demo cfg/yolo-face.data cfg/yolo-face.cfg backup/yolo-face_final.weights

• train
• deploy

• Intel Core i7, 2.5 GHz
• 1 Processor, 4 Cores
• 16GB MEMORY
• AMD Radeon R9 M370X

• Credits: https://github.com/AlexeyAB/darknet, https://github.com/xhuvom/darknetFaceID
• Other References: https://github.com/quanhua92/darknet
• For more information see the Darknet project website.
• For questions or issues please use the Google Group.