Research map / Detection and segmentation / CNN features
$\alpha$-IoU: A Family of Power Intersection over Union Losses for Bounding Box Regression
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 |
|---|---|---|---|
| power α=3 on IoU loss | L_IoU YOLOv5s · PASCAL VOC | mAP75:95: 32.45 -> 34.46 | Helped |
| power α=3 on IoU loss | L_IoU YOLOv5s · PASCAL VOC | AP50: 78.81 -> 78.62 | No clear effect |
| power α=3 on IoU loss | L_IoU YOLOv5x · MS COCO | mAP: 48.42 -> 48.67 | Helped |
| power α=3 on IoU loss | L_IoU DETR · MS COCO | mAP: 39.23 -> 40.18 | Helped |
| power α=3 on DIoU loss | L_DIoU Faster R-CNN · MS COCO | mAP | Helped |
| power α=3 on IoU loss under bbox noise η=0.3 | L_IoU YOLOv5s · PASCAL VOC noisy | mAP75:95: 3.54 -> 4.29 | Helped |
| α=10 instead of α=3 (ablation) | α-IoU YOLOv5s · PASCAL VOC | mAP | Hurt |
| α in (0,1) (ablation) | α-IoU · PASCAL VOC | mAP | Hurt |
| vary α2 for penalty term (0.5, 1, 3) (ablation) | α-GIoU YOLOv5s · PASCAL VOC | mAP: 53.75 / 53.29 / 53.44 | No clear effect |
| power α=3 on IoU loss with 25% training data | L_IoU YOLOv5s · 25% PASCAL VOC | mAP75:95: 16.13 -> 17.57 | Helped |
Closest papers by meaning
- Distance-IoU Loss: Faster and Better Learning for Bounding Box Regression (AAAI 2020 · similarity 0.92)
- Generalized Intersection Over Union: A Metric and a Loss for Bounding Box Regression (CVPR 2019 · similarity 0.91)
- Bounding Box Stability against Feature Dropout Reflects Detector Generalization across Environments (ICLR 2024 · similarity 0.85)
- Union-over-Intersections: Object Detection beyond Winner-Takes-All (ICLR 2025 · similarity 0.85)
- Theoretically Achieving Continuous Representation of Oriented Bounding Boxes (CVPR 2024 · similarity 0.85)
- Rethinking Boundary Discontinuity Problem for Oriented Object Detection (CVPR 2024 · similarity 0.84)
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.