darknet
darknet copied to clipboard
YOLOV2 ANCHORS
yolov3 and yolov2 anchors are quite different, like yolov3 anchors int integer values where yolov2 are in floating points that's because yolov2 anchors are divided by the height and width of the network. I tried to train a yolov2-tiny model using your repository. using your repo I got anchors 17,29 29,21. what should I do to train the yolov2-tiny network? should I use these anchors or I should divide these with height and width in yolov2 format? I will use this your repo for yolov2 training.