Research map / Detection and segmentation / CNN features
Learning Modulated Loss for Rotated Object Detection
Read the paperCheck 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 smooth-ℓ1 with modulated rotation loss in five-parameter regression (ablation) | RetinaNet-H with ResNet50 · DOTA | mAP: 62.14 -> 64.49 | Helped |
| Replace five-parameter with eight-parameter regression using smooth-ℓ1 (ablation) | RetinaNet-H with ResNet50 · DOTA | mAP: 62.14 -> 65.59 | Helped |
| Use modulated rotation loss with eight-parameter regression (ablation) | RetinaNet-H with ResNet50 · DOTA | mAP: 65.59 -> 66.77 | Helped |
| Expand the five-parameter angle range from [-π/2,0) to [-π,0) (ablation) | RetinaNet-H · DOTA | mAP: 62.14 -> 62.39 | Helped |
| Use tan angular encoding with smooth-ℓ1 (ablation) | RetinaNet-H · DOTA | mAP: NAN | Hurt |
| Add data augmentation (ablation) | RSDet with ResNet50 · DOTA | mAP: 66.77 -> 70.79 | Helped |
| Add class balancing by copying underrepresented categories (ablation) | RSDet with ResNet50 and data augmentation · DOTA | mAP: 70.79 -> 71.22 | Helped |
| Use ResNet101 instead of ResNet50 (ablation) | RSDet with augmentation and class balancing · DOTA | mAP: 71.22 -> 72.16 | Helped |
| Use ResNet152 instead of ResNet101 (ablation) | RSDet with augmentation and class balancing · DOTA | mAP: 72.16 -> 73.51 | Helped |
| Add a refinement stage (ablation) | RSDet with ResNet152 · DOTA | mAP: 73.51 -> 74.13 | Helped |
| Apply modulated rotation loss and eight-parameter regression (ablation) | RetinaNet-H with ResNet152 · ICDAR2015 | performance: 76.8 -> 83.2 | Helped |
| Apply modulated rotation loss and eight-parameter regression (ablation) | RetinaNet-H with ResNet152 · HRSC2016 | performance: 82.4 -> 86.5 | Helped |
Closest papers by meaning
- Learning High-Precision Bounding Box for Rotated Object Detection via Kullback-Leibler Divergence (NeurIPS 2021 · similarity 0.89)
- Rethinking Boundary Discontinuity Problem for Oriented Object Detection (CVPR 2024 · similarity 0.87)
- ReDiffDet: Rotation-equivariant Diffusion Model for Oriented Object Detection (CVPR 2025 · similarity 0.84)
- FRED: Towards a Full Rotation-Equivariance in Aerial Image Object Detection (AAAI 2024 · similarity 0.83)
- Theoretically Achieving Continuous Representation of Oriented Bounding Boxes (CVPR 2024 · similarity 0.83)
- R2Det: Exploring Relaxed Rotation Equivariance in 2D Object Detection (ICLR 2025 · similarity 0.83)
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.