pytorch-cpn
pytorch-cpn copied to clipboard
some question about the human detector.
in this paper, I see that CPN used the image cropped in raw image with the bounding box of FPN as input. Where is the FPN in your reimplement ?