Research map / Detection and segmentation / CNN features
Adaptive Rotated Convolution for Rotated Object Detection
Yifan Pu, Yiru Wang, Zhuofan Xia, Yizeng Han, Yulin Wang, Weihao Gan, Zidong Wang, Shiji Song, Gao Huang
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 changed | Where | Measured | Effect |
|---|---|---|---|
| replace static convolutions with adaptive rotated convolution (ARC) in the backbone | Oriented R-CNN ResNet-50 · DOTA-v1.0 single-scale | mAP: 75.81 -> 77.35 | Improved |
| use ARC backbone across detectors | Rotated RetinaNet, R3Det, S2ANet, Rotated Faster R-CNN, CFA · DOTA-v1.0 | mAP: +3.03, +2.62, +1.36, +1.60, +4.16 mAP | Improved |
| use ARC backbone across detectors on HRSC2016 | Rotated RetinaNet, S2ANet, Oriented R-CNN · HRSC2016 | mAP: +1.27, +2.12, +1.84 mAP | Improved |
| add adaptive kernel rotation versus static convolution (ablation) | Oriented R-CNN · DOTA-v1.0 | mAP: 75.81 -> 76.97 | Improved |
| increase number of kernels from 1 to 4 (conditional computation) (ablation) | ARC-R50 · DOTA-v1.0 | mAP: 76.97 -> 77.35 | Improved |
| increase number of kernels from 4 to 6 (ablation) | ARC-R50 · DOTA-v1.0 | mAP: 77.35 -> 77.38 | No clear effect |
| replace convolutions in more backbone stages with ARC (ablation) | ARC-R50 · DOTA-v1.0 | mAP: 77.17 -> 77.35 | Improved |
| add spatial information encoding in routing function (ablation) | ARC routing function · DOTA-v1.0 | mAP: 76.41 -> 76.80 | Improved |
| add adaptive combination of kernel weights in routing function (ablation) | ARC routing function · DOTA-v1.0 | mAP: 76.41 -> 76.98 | Improved |
| use ARC with four kernels instead of static ResNet-50 | Oriented R-CNN · DOTA-v1.0 inference speed | FPS: 29.9 -> 29.2 | Hurt |
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.
- ReDet — ReDet: A Rotation-Equivariant Detector for Aerial Object Detection (CVPR 2021) · better on 1 of 1 shared result
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
- GRA: Detecting Oriented Objects through Group-wise Rotating and Attention (ECCV 2024 · similarity 0.91)
- Oriented R-CNN for Object Detection (ICCV 2021 · similarity 0.89)
- Learning High-Precision Bounding Box for Rotated Object Detection via Kullback-Leibler Divergence (NeurIPS 2021 · similarity 0.88)
- Weakly Supervised Rotation-Invariant Aerial Object Detection Network (CVPR 2022 · similarity 0.87)
- FRED: Towards a Full Rotation-Equivariance in Aerial Image Object Detection (AAAI 2024 · similarity 0.87)
- ReDiffDet: Rotation-equivariant Diffusion Model for Oriented Object Detection (CVPR 2025 · similarity 0.87)
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.
