0

I have a short question. I understand the concept of RPN but one small details keeps me from implementing it. How should I design the training loop given that I have to use only a subset of anchor boxes (128 positive and 128 negative). In other words, if the ouput of reg layer is a map of every bounding box, how do I update only the bounding boxes that match my 128 positive/128 negative samples

Bogdan Duduc
  • 11
  • 1
  • Hello. Could you please put your **specific** question in the title? "region proposal network training" is **not** a question and it's very general/vague. Moreover, what do you mean by "reg layer"? – nbro Apr 27 '21 at 23:27
  • Ok,i will update the title. Reg layer is regressive layer for bounding box coordinates – Bogdan Duduc May 08 '21 at 20:32

0 Answers0