AI Paper Atlas

Research map / Detection and segmentation / CNN features

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

Shaoqing Ren, Kaiming He, Ross Girshick, Jian Sun

TPAMI 2016 · Detection and segmentation › CNN features · approach: Region-based and query-based detectors · 74,198 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
share conv layers between RPN and Fast R-CNN (ablation)Faster R-CNN ZF · VOC07 testmAP: 58.7 -> 59.9Improved
add cls scores for ranking proposals (vs random sampling) at N=100 (ablation)RPN+ZF · VOC07 testmAP: 44.6 -> 55.1Improved
add reg layer (vs anchor boxes as proposals) (ablation)RPN+ZF · VOC07 testmAP: 52.1 -> 56.8Improved
apply NMS to proposals (ablation)RPN+ZF · VOC07 testmAP: 55.2 -> 56.8No clear effect
replace ZF with VGG-16 in RPN (ablation)RPN · VOC07 test (SS-trained ZF detector)mAP: 56.8 -> 59.2Improved
replace SS with RPN in train and test (ablation)Fast R-CNN ZF · VOC07 testmAP: 58.7 -> 59.9Improved
use shared features for VGG-16 (ablation)Faster R-CNN VGG-16 · VOC07 testmAP: 68.5 -> 69.9Improved
add VOC12 trainval to trainingFaster R-CNN VGG-16 · VOC07 testmAP: 69.9 -> 73.2Improved
add COCO pretraining to VOC07+12Faster R-CNN VGG-16 · VOC07 testmAP: 73.2 -> 78.8Improved
add multiple scales (3 scales, 1 ratio vs 1 scale 1 ratio) (ablation)Faster R-CNN VGG-16 · VOC07 testmAP: 66.7 -> 69.8Improved
add multiple aspect ratios (3 ratios, 128^2 scale) (ablation)Faster R-CNN VGG-16 · VOC07 testmAP: 65.8 -> 68.8Improved
add 3 aspect ratios to 3 scales (ablation)Faster R-CNN VGG-16 · VOC07 testmAP: 69.8 -> 69.9No clear effect

Later papers that compared with this method

175 later accepted papers report Faster R-CNN as a baseline in their results tables; 164 of them report a better number on most of the results they share with it.

167 more not listed.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy