AI Paper Atlas

Research map / Detection and segmentation / CNN features

Oriented R-CNN for Object Detection

Xingxing Xie, Gong Cheng, Jiabao Wang, Xiwen Yao, Junwei Han

ICCV 2021 · Detection and segmentation › CNN features · approach: Region-based and query-based detectors · 1,549 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
replace RoI Transformer+ proposal generation with oriented RPN (Oriented R-CNN)two-stage oriented detector, R-50-FPN · DOTA testmAP: 74.61 -> 75.87Improved
replace RoI Transformer+ proposal generation with oriented RPN (Oriented R-CNN)two-stage oriented detector, R-50-FPN · DOTA speed (RTX 2080Ti)FPS: 11.3 -> 15.1Improved
use oriented R-CNN instead of one-stage RetinaNet-OR-50-FPN detector · DOTA testmAP / FPS: 68.43 -> 75.87 mAP; 16.1 -> 15.1 FPSMixed
add multi-scale training and testingOriented R-CNN R-50-FPN · DOTA testmAP: 75.87 -> 80.87Improved
add multi-scale training and testingOriented R-CNN R-101-FPN · DOTA testmAP: 76.28 -> 80.52Improved
use ResNet-101-FPN instead of ResNet-50-FPN backboneOriented R-CNN · DOTA testmAP: 75.87 -> 76.28Improved
use ResNet-101-FPN instead of ResNet-50-FPN backboneOriented R-CNN · HRSC2016 (VOC12 metric)mAP: 96.50 -> 97.60Improved
use ResNet-101-FPN instead of ResNet-50-FPN backboneOriented R-CNN · HRSC2016 (VOC07 metric)mAP: 90.40 -> 90.50No clear effect
reduce proposals from 2000 to 1000 per image patch (ablation)oriented RPN · DOTA valrecall: 92.80 -> 92.20Hurt
reduce proposals from 1000 to 300 per image patch (ablation)oriented RPN · DOTA valrecall: 92.20 -> 81.60Hurt
outperform all R-101-FPN comparison methods using only R-50-FPNOriented R-CNN · DOTA testmAP: 75.87 vs 75.02 (Gliding Vertex, R-101-FPN)Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

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