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 combine scores of stage1 and stage2 #16

Open emergencyd opened 5 years ago

emergencyd commented 5 years ago

What if their prediction class is different? Like stage1 predict the box as cat, while stage2 predict the box as dog. Thank you

bowenc0221 commented 5 years ago

We multiply the softmax score elementwise, which is a 21-D vector for pascal voc and 81-D vector for coco. For example, stage 1 predict [0.9, 0.1] and stage 2 predict [0.2, 0.8], the final score is [0.18, 0.08].