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

Checking my understanding for the intuition #20

Open ydixon opened 3 years ago

ydixon commented 3 years ago

So is the whole DCR setup similar to setting up Faster RCNN + classifier?

For DCR v1, is it just letting the classifier post-process the output from the faster RCNN to correct the results? And to train the classifier, we get data samples by GT patches, TP prediction patches (From Faster RCNN), FP prediction patches (From Faster RCNN)?

For DCR v2, speed is improved by stealing some low level features layers instead of having a completely new branch of CNN layers?