Skip to content

Latest commit

 

History

History
 
 

rpn

RPN

Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks

Abstract

State-of-the-art object detection networks depend on region proposal algorithms to hypothesize object locations. Advances like SPPnet and Fast R-CNN have reduced the running time of these detection networks, exposing region proposal computation as a bottleneck. In this work, we introduce a Region Proposal Network (RPN) that shares full-image convolutional features with the detection network, thus enabling nearly cost-free region proposals. An RPN is a fully convolutional network that simultaneously predicts object bounds and objectness scores at each position. The RPN is trained end-to-end to generate high-quality region proposals, which are used by Fast R-CNN for detection. We further merge RPN and Fast R-CNN into a single network by sharing their convolutional features---using the recently popular terminology of neural networks with 'attention' mechanisms, the RPN component tells the unified network where to look. For the very deep VGG-16 model, our detection system has a frame rate of 5fps (including all steps) on a GPU, while achieving state-of-the-art object detection accuracy on PASCAL VOC 2007, 2012, and MS COCO datasets with only 300 proposals per image. In ILSVRC and COCO 2015 competitions, Faster R-CNN and RPN are the foundations of the 1st-place winning entries in several tracks.

Results and Models

Backbone Style Lr schd Mem (GB) Inf time (fps) AR1000 Config Download
R-50-FPN caffe 1x 3.5 22.6 58.7 config model | log
R-50-FPN pytorch 1x 3.8 22.3 58.2 config model | log
R-50-FPN pytorch 2x - - 58.6 config model | log
R-101-FPN caffe 1x 5.4 17.3 60.0 config model | log
R-101-FPN pytorch 1x 5.8 16.5 59.7 config model | log
R-101-FPN pytorch 2x - - 60.2 config model | log
X-101-32x4d-FPN pytorch 1x 7.0 13.0 60.6 config model | log
X-101-32x4d-FPN pytorch 2x - - 61.1 config model | log
X-101-64x4d-FPN pytorch 1x 10.1 9.1 61.0 config model | log
X-101-64x4d-FPN pytorch 2x - - 61.5 config model | log

Citation

@inproceedings{ren2015faster,
  title={Faster r-cnn: Towards real-time object detection with region proposal networks},
  author={Ren, Shaoqing and He, Kaiming and Girshick, Ross and Sun, Jian},
  booktitle={Advances in neural information processing systems},
  year={2015}
}