Research map / Detection and segmentation / CNN features
RiO-DETR: DETR for Real-time Oriented Object Detection
Zhangchi Hu, Yifan Zhao, Yansong Peng, Wenzhang SUN, Xiangchen Yin, Jie Chen, Peixi Wu, Hebei Li, xinghao wang, Dongsheng Jiang, Xiaoyan Sun
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 |
|---|---|---|---|
| add Hausdorff matching and KLD loss (ablation) | Oriented RT-DETRv2 · DIOR-R (RiO-DETR-m) | AP50: 70.35 -> 72.86 | Improved |
| add D-FINE universal matching (ablation) | Oriented RT-DETRv2 + Hausdorff/KLD · DIOR-R (RiO-DETR-m) | AP50: 72.86 -> 73.33 | Improved |
| add Dense O2O (ablation) | baseline with universal matching · DIOR-R (RiO-DETR-m) | AP50: 73.33 -> 73.47 | Improved |
| add Geometry-Decoupled Query Encoding (ablation) | implemented baseline · DIOR-R (RiO-DETR-m) | AP50: 73.47 -> 74.18 | Improved |
| add Rotation-Rectified Orthogonal Attention (ablation) | baseline + decoupled query · DIOR-R (RiO-DETR-m) | AP50: 74.18 -> 74.74 | Improved |
| add Decoupled Periodic Refinement (ablation) | baseline + content-driven angle · DIOR-R (RiO-DETR-m) | AP50: 74.74 -> 75.46 | Improved |
| add Oriented Dense O2O (ablation) | RiO-DETR-m without it · DIOR-R (RiO-DETR-m) | AP50: 75.46 -> 75.73 | Improved |
| encode angle in positional query in addition to center (ablation) | query encoding with center only · DIOR-R (RiO-DETR-m) | AP50: 72.56 -> 72.34 | Hurt |
| encode size (w,h) in query (ablation) | center-only query · DIOR-R (RiO-DETR-m) | AP50: 72.56 -> 73.81 | Improved |
| add angle to center+size query encoding (ablation) | query encoding center+size · DIOR-R (RiO-DETR-m) | AP50: 73.81 -> 73.57 | Hurt |
| add shortest-path periodic L1 loss only (ablation) | refinement without periodic components · DIOR-R (RiO-DETR-m) | AP50: 74.18 -> 74.32 | Improved |
| add periodic update only (no SP-L1) (ablation) | refinement without periodic components · DIOR-R (RiO-DETR-m) | AP50: 74.18 -> 74.05 | Hurt |
Closest papers by meaning
- Projecting Points to Axes: Oriented Object Detection via Point-Axis Representation (ECCV 2024 · similarity 0.88)
- DETRs Beat YOLOs on Real-time Object Detection (CVPR 2024 · similarity 0.88)
- Accelerating DETR Convergence via Semantic-Aligned Matching (CVPR 2022 · similarity 0.87)
- MI-DETR: An Object Detection Model with Multi-time Inquiries Mechanism (CVPR 2025 · similarity 0.87)
- Rethinking Boundary Discontinuity Problem for Oriented Object Detection (CVPR 2024 · similarity 0.86)
- ReDiffDet: Rotation-equivariant Diffusion Model for Oriented Object Detection (CVPR 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.
