Replies: 1 comment
-
Fast R-CNN takes an input image and region proposals as input, as described in the original paper. So you need the RPN in front of Fast R-CNN as its output is part of the input of Fast R-CNN.. this is also noted in the respective Git. The RPNHead is here and a default config as used in the later Faster R-CNN can be found in its config. |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
I hope to replicate the original version of fast rcnn. I noticed that the replicated version of fast rcnn of mmdetection includes Shared2FCBBoxHead, which is not what I want.
I see an RPN detector in mmdetection. It seems that there is a Shared2FCBBoxHead less than the fast rcnn version in the network structure. Is this the corresponding original fast rcnn?
In short, I hope to find a version of mmdetection similar to the original version of fast rcnn. Can anyone help me?
Beta Was this translation helpful? Give feedback.
All reactions