search

yangxue0827 / R2CNN_FPN_Tensorflow

R2CNN: Rotational Region CNN Based on FPN (Tensorflow)
419 stars 139 forks source link

The classification loss seems not to be added into the total_loss? #62

Closed Remember2018 closed 5 years ago

Remember2018 commented 5 years ago

Hi Dear Yang, thanks a lot for you to share the code. I read the code and found that in the faster_rcnn module, the fast_rcnn classificatoin loss was not added into the slim losses. https://github.com/yangxue0827/R2CNN_FPN_Tensorflow/blob/0dc1b201e39f4e564afe11cb72b78ae849d61e80/libs/fast_rcnn/build_fast_rcnn1.py#L313

And only the box losses were added. If convenient, could you please give some reasons? Thanks a lot!

yangxue0827 commented 5 years ago

https://github.com/yangxue0827/R2CNN_FPN_Tensorflow/blob/0dc1b201e39f4e564afe11cb72b78ae849d61e80/tools/train1.py#L135 @Remember2018

Recommended use https://github.com/DetectionTeamUCAS/R2CNN_Faster-RCNN_Tensorflow

Remember2018 commented 5 years ago

@yangxue0827 Thanks a lot for your kind advice! Yes! I'm using the R2CNN_Faster-RCNN_Tensorflow and has obtained some good results. I'm trying to integrate the FPN module, and found your R2CNN_FPN_Tensorflow. Thanks again for sharing your code!

yangxue0827 commented 5 years ago

https://github.com/DetectionTeamUCAS/R2CNN_Faster-RCNN_Tensorflow https://github.com/DetectionTeamUCAS/FPN_Tensorflow @Remember2018

Remember2018 commented 5 years ago

@yangxue0827 Thanks a lot!