Research map / Detection and segmentation / CNN features
ReDet: A Rotation-Equivariant Detector for Aerial Object Detection
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the ordinary backbone with a C8 rotation-equivariant backbone (ablation) | R50-FPN detector · DOTA-v1.5 test | mAP: 65.03 -> 66.86 | Helped |
| Increase the backbone rotation group from C4 to C8 (ablation) | ReR50-ReFPN · DOTA-v1.5 test | mAP: 65.43 -> 66.86 | Helped |
| Increase the backbone rotation group from C8 to C16 (ablation) | ReR50-ReFPN · DOTA-v1.5 test | mAP: 66.86 -> 64.36 | Hurt |
| Add orientation alignment to RoI feature warping (ablation) | ReDet with ReR50-ReFPN · DOTA-v1.5 test | mAP: 65.99 -> 66.86 | Helped |
| Replace orientation alignment with orientation max pooling (ablation) | ReDet with ReR50-ReFPN · DOTA-v1.5 test | mAP: 65.99 -> 64.60 (-1.39) | Hurt |
| Use 1 orientation channel for RiRoI Align interpolation (ablation) | ReDet · DOTA-v1.5 test | mAP: 65.99 -> 66.44 (+0.45) | Helped |
| Increase RiRoI Align interpolation from 2 to 4 channels (ablation) | ReDet · DOTA-v1.5 test | mAP: 66.86 -> 66.32 | Hurt |
| Use ReDet without rotation augmentation instead of a rotation-augmented baseline (ablation) | Detector with R18/ReR18 backbone, 1x schedule · DOTA-v1.5 test | mAP: 64.07 -> 62.62 | Hurt |
| Use parameter-matched ReDet instead of the rotation-augmented baseline (ablation) | Detector with R18/ReR18 backbone, 1x schedule · DOTA-v1.5 test | mAP: 64.07 -> 66.66; 2.59 mAP improvements | Helped |
| Replace baseline with ReDet | Oriented detector · DOTA-v1.0 OBB, single-scale | mAP: 74.12 -> 76.25 | Helped |
| Replace baseline with ReDet | Oriented detector · DOTA-v1.5 OBB, multi-scale | mAP: 74.90 -> 76.80 | Helped |
| Replace baseline with ReDet | Oriented detector · HRSC2016, VOC2007 | mAP: 90.17 -> 90.46 | Helped |
Closest papers by meaning
- FRED: Towards a Full Rotation-Equivariance in Aerial Image Object Detection (AAAI 2024 · similarity 0.94)
- R2Det: Exploring Relaxed Rotation Equivariance in 2D Object Detection (ICLR 2025 · similarity 0.88)
- ReDiffDet: Rotation-equivariant Diffusion Model for Oriented Object Detection (CVPR 2025 · similarity 0.87)
- RemDet: Rethinking Efficient Model Design for UAV Object Detection (AAAI 2025 · similarity 0.86)
- ABBSPO: Adaptive Bounding Box Scaling and Symmetric Prior based Orientation Prediction for Detecting Aerial Image Objects (CVPR 2025 · similarity 0.86)
- FBRT-YOLO: Faster and Better for Real-Time Aerial Image Detection (AAAI 2025 · similarity 0.86)
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.