Open NEELMCW opened 5 years ago
@NEELMCW Hi,
You can visualize relative sizes of objects as cloud of points (each point(x,y) = each object size(w,h)) after that each image is resized to the 416x416:
./darknet detector calc_anchors obj.data -num_of_clusters 9 -width 416 -height 416 -show
You can change this code: https://github.com/AlexeyAB/darknet/blob/5343aa423563c107e0071b1427ad5defc27b56d2/src/detector.c#L1028-L1216 More: https://github.com/AlexeyAB/darknet/issues/2186#issuecomment-453672469
Thanks @AlexeyAB
Hello @AlexeyAB
Is there any snippet that dumps the object size distributions for a given train set annotations in yolo txt format ?