SHI-Labs / Decoupled-Classification-Refinement

Revisiting RCNN: On Awakening the Classification Power of Faster RCNN (ECCV 2018)
https://arxiv.org/abs/1803.06799
MIT License
167 stars 24 forks source link

How to sample boxes in the DCRV2? #19

Open HuangLian126 opened 4 years ago

HuangLian126 commented 4 years ago

Thanks for your nice work, however I have some questions: 1)In the DCRV2, paepr writes that "During training, we first sample R boxes from the outputs of the bounding box regression branch in the base detector, then we use these boxes as ROIs to extract regional features in the DCR V2 module. " You mean these boxes are final output of base detector, not the propsals of RPN? 2)Do you use the "top-sampling" that samples top 50% confidence scores of boxes? In other words, you do not select boxes with the IOU? 3)Do you sample these boxes before softmax?