Closed junehyeok129 closed 3 months ago
@junehyeok129 Thank you for the comment!
In RCNN, an existing non-deep learning-based region proposal method, such as Selective Search, is used, not a segmentation method. In Faster R-CNN, to address the shortcomings of using existing region proposals, a Region Proposal Network (RPN) is introduced, which can be jointly trained with the detection component.
I learned that you have to do segmentation in order to do the Region proposal in R-CNN. However, have to do so much segmentation that there is a resource consumption. Is there any other good way?