Research map / Detection and segmentation / CNN features
Learning a Unified Sample Weighting Network for 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 |
|---|---|---|---|
| Add SWN sample weighting | Faster R-CNN, Mask R-CNN, and RetinaNet · COCO test-dev | AP: 36.7 -> 38.5; 40.3 -> 41.4; 37.5 -> 39.0; 41.4 -> 42.5; 35.9 -> 37.2; 39.0 -> 40.8 | Helped |
| Add SWN sample weighting | Faster R-CNN and RetinaNet · VOC2007 test | AP: 51.0 -> 52.5; 54.2 -> 56.0; 52.0 -> 53.4; 55.3 -> 56.8 | Helped |
| Add SWN | Faster R-CNN + ResNet-50 · COCO val2017 | AP: 36.4 -> 38.2 | Helped |
| Add Soft-NMS to SWN | Faster R-CNN + ResNet-50 · COCO val2017 | AP: 38.2 -> 39.2 | Helped |
| Use sample weighting for classification only (ablation) | Faster R-CNN + ResNet-50 · COCO val2017 | AP: 36.4 -> 36.7 | Helped |
| Use sample weighting for regression only (ablation) | Faster R-CNN + ResNet-50 · COCO val2017 | AP: 36.4 -> 37.0 | Helped |
| Add joint classification and regression weighting rather than no task weighting (ablation) | Faster R-CNN + ResNet-50 · COCO val2017 | AP: 36.4 -> 38.2 | Helped |
| Set the shared regularization λ to 0.1 instead of 0.5 (ablation) | SWN · COCO val2017 | AP: 38.2 -> 29.3 | Hurt |
| Set the shared regularization λ to 0.3 instead of 0.5 (ablation) | SWN · COCO val2017 | AP: 38.2 -> 37.4 | Hurt |
| Set the shared regularization λ to 0.7 instead of 0.5 (ablation) | SWN · COCO val2017 | AP: 38.2 -> 37.9 | Hurt |
| Set the shared regularization λ to 1.0 instead of 0.5 (ablation) | SWN · COCO val2017 | AP: 38.2 -> 37.2 | Hurt |
Closest papers by meaning
- A Dual Weighting Label Assignment Scheme for Object Detection (CVPR 2022 · similarity 0.92)
- Unbalanced Optimal Transport: A Unified Framework for Object Detection (CVPR 2023 · similarity 0.87)
- RankDetNet: Delving Into Ranking Constraints for Object Detection (CVPR 2021 · similarity 0.87)
- R-FCN: Object Detection via Region-based Fully Convolutional Networks (NeurIPS 2016 · similarity 0.86)
- Towards Human-Machine Cooperation: Self-Supervised Sample Mining for Object Detection (CVPR 2018 · similarity 0.86)
- Sampling Techniques for Large-Scale Object Detection From Sparsely Annotated Objects (CVPR 2019 · 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.