An Implementation of Faster RCNN with Study for Region Sampling
read the original abstract
We adapted the join-training scheme of Faster RCNN framework from Caffe to TensorFlow as a baseline implementation for object detection. Our code is made publicly available. This report documents the simplifications made to the original pipeline, with justifications from ablation analysis on both PASCAL VOC 2007 and COCO 2014. We further investigated the role of non-maximal suppression (NMS) in selecting regions-of-interest (RoIs) for region classification, and found that a biased sampling toward small regions helps performance and can achieve on-par mAP to NMS-based sampling when converged sufficiently.
This paper has not been read by Pith yet.
Forward citations
Cited by 1 Pith paper
-
Automatic Colon Polyp Detection using Region based Deep CNN and Post Learning Approaches
Region-based deep CNN with transfer learning and post-learning methods achieves better polyp detection performance than prior systems on large colonoscopy image and video databases.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.