pythonobject-detectionyoloyolov4

What does yolov4.conv.137 mean in Yolov4 darknet?

I can't find a description of these weights used in tutorial for training Yolov4 darknet and I can't really understand where they came from, COCO? PASCAL VOC? ImageNet? I can't understand also if running the command !./darknet detector train ./obj.data ./yolov4-obj.cfg ./yolov4.conv.137 -dont_show -map means fine-tuning (i.e. simply initializing the model with weights without freezing anything) or transfer learning (freezing layers in the model)??

Solution

  • They are learned from MS COCO: ref