search

krumo / Domain-Adaptive-Faster-RCNN-PyTorch

Domain Adaptive Faster R-CNN in PyTorch
MIT License
305 stars 68 forks source link

Image level adaptation with FPN #26

Closed gabridego closed 3 years ago

gabridego commented 3 years ago

Hello @krumo , thanks for the great work! I am trying to use your project with a ResNet backbone with a feature pyramid network on top, I would like to ask you if you have any clue about how to apply the domain discriminator at the image level in this case, as the backbone returns feature maps of different sizes, and it is not possible to simply concatenate them. I've noticed that it is stated that DomainAdaptationModule works for both FPN and non-FPN and that a boolean USE_FPN parameter is mentioned but not used, maybe you already thought of an expected behavior? https://github.com/krumo/Domain-Adaptive-Faster-RCNN-PyTorch/blob/2f7892d1a22b0ba53f1dd807da15be36a7bbfea5/maskrcnn_benchmark/modeling/da_heads/da_heads.py#L15-L21 Thanks!

krumo commented 3 years ago

Hi, thanks for your interest! Yes, we tried combining the domain adaptation module with FPN, the design and the experimental results have been released in our IJCV paper Scale-Aware Domain Adaptive Faster R-CNN. You could also check our implementation in sa-da-faster.