regression target - Githubissues

AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )

http://pjreddie.com/darknet/

Other

21.75k stars 7.96k forks source link

regression target #834

Open kwkwvenusgod opened 6 years ago

kwkwvenusgod commented 6 years ago

As we all know, generating proper anchor boxed and feeding them to the network is really important step in this object detection. But from this impl https://github.com/thtrieu/darkflow/blob/master/darkflow/net/yolov2/data.py#L21, it seems to omit the anchor scheme. So I jumped to your implementation. Can I know which part is to implement the network feed correlated with the anchor boxes? I appreciate it a lot if you can answer me.

AlexeyAB commented 6 years ago

For Yolov2: biases[] - is an anchor boxes:

Detection - get bounded boxes: https://github.com/AlexeyAB/darknet/blob/573d7e80814a4cc3c08897f6c0f67ea189339856/src/region_layer.c#L78-L90
Training - get delta (and later calculating loss): https://github.com/AlexeyAB/darknet/blob/573d7e80814a4cc3c08897f6c0f67ea189339856/src/region_layer.c#L92-L111