search

allanzelener / YAD2K

YAD2K: Yet Another Darknet 2 Keras
Other
2.71k stars 878 forks source link

Using weight from darknet #79

Open Bala93 opened 6 years ago

Bala93 commented 6 years ago

After training using darknet framework. Can the obtained weights be used directly in the testing stage. If that is the case . How it should be done ? And this is for custom dataset with single class.

ACfromecs commented 6 years ago

@Bala93 example: use python3, training with darknet/cfg/tiny-yolo-voc.cfg, get weight file darknet/resoults/tiny-yolo-voc.backup, classesname file darknet/data/voc.data: step 1: cp darknet/resoults/tiny-yolo-voc.backup tiny-yolo-voc.weights step 2: cp darknet/data/voc.data voc.txt step 3: python3 yad2k.py darknet/cfg/tiny-yolo-voc.cfg tiny-yolo-voc.weights model_data/tiny-yolo-voc.h5 step 4: python3 test_yolo.py model_data/tiny-yolo-voc.h5 --classes_path voc.txt --anchors_path model_data/tiny-yolo-voc_anchors.txt