doubt in faster rcnn config #5470
Unanswered
zhaoxin111
asked this question in
Q&A
Replies: 1 comment 1 reply
-
The five feature maps are used for proposal generation. But for RoI feature extraction we only use the 4 feature maps. |
Beta Was this translation helpful? Give feedback.
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
According to the logic of the _bbox_forward method, only the first four feature layers are responsible for bbox feature representation. I check the config of faster rcnn in detectron2, all 5 feature maps are responsible for the proposal feature representation. Why mmdet only use 4 feature maps in bbox_roi_extractor?
Beta Was this translation helpful? Give feedback.
All reactions