search

hyz-xmaster / VarifocalNet

VarifocalNet: An IoU-aware Dense Object Detector
Apache License 2.0
346 stars 52 forks source link

Questions about reimplementing the experiments in your paper. #15

Open rainylt opened 3 years ago

rainylt commented 3 years ago

Thanks for your great contribution. I have read your paper, but have some questions about the experiments in Table 1.

  1. The replacement of the predicted results is only implemented in the inference process rather than training?

  2. How to replace the predicted classification score with its ground truth, In other words, how to assign labels to anchors when replacing? Is it the same way in FCOS, just assign the anchor to the gt box it's in?

hyz-xmaster commented 3 years ago

Hi @lt1103725556, thank you for interesting our paper. Please see this answer. If you still have problems after reading that, please let me know.

rainylt commented 3 years ago

Hi @hyz-xmaster, thanks for your early reply. I have seen your answer. Although I haven't still grasped your code, I can have a shallow understanding that you use the training phase just for getting the corresponding ground truth easily, so setting the overall lr = 0.0 is ok?

hyz-xmaster commented 3 years ago

I think setting lr = 0.0 should probably work, although I haven't tried. Just have a go.