Research map / Detection and segmentation / CNN features
What Makes for End-to-End 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 one-to-many with one-to-one assignment (ablation) | RetinaNet · COCO val2017 | AP: 7.7 -> 33.6 | Helped |
| replace one-to-many with one-to-one assignment (ablation) | CenterNet · COCO val2017 | AP: 24.9 -> 23.4 | Hurt |
| replace one-to-many with one-to-one assignment (ablation) | FCOS · COCO val2017 | AP: 17.3 -> 34.9 | Helped |
| add classification cost to one-to-one assignment (ablation) | RetinaNet · COCO val2017 | AP: 33.6 -> 36.0 | Helped |
| add classification cost to one-to-one assignment (ablation) | CenterNet · COCO val2017 | AP: 23.4 -> 33.3 | Helped |
| add classification cost to one-to-one assignment (ablation) | FCOS · COCO val2017 | AP: 34.9 -> 35.9 | Helped |
| add classification cost to matching (ablation) | DETR · COCO val2017 | AP: 12.6 -> 40.0 | Helped |
| add classification cost to matching (ablation) | Deformable DETR · COCO val2017 | AP: 12.0 -> 44.0 | Helped |
| add classification cost to matching (ablation) | Sparse R-CNN · COCO val2017 | AP: 20.1 -> 45.0 | Helped |
| use predicted instead of predefined location cost (ablation) | FCOS + one-to-one + classification cost · COCO val2017 | AP: 35.9 -> 38.9 | Helped |
| use predicted instead of predefined location cost (ablation) | RetinaNet + one-to-one + classification cost · COCO val2017 | AP: 36.0 -> 37.5 | Helped |
| use one-to-one assignment and classification cost without NMS | RetinaNet · CrowdHuman validation | Recall: 88.6 -> 98.1 | Helped |
Closest papers by meaning
- End-to-End Object Detection With Fully Convolutional Network (CVPR 2021 · similarity 0.89)
- Dense Distinct Query for End-to-End Object Detection (CVPR 2023 · similarity 0.87)
- One-to-Few Label Assignment for End-to-End Dense Detection (CVPR 2023 · similarity 0.87)
- Integrating Diverse Assignment Strategies into DETRs (AAAI 2026 · similarity 0.86)
- DETRs With Hybrid Matching (CVPR 2023 · similarity 0.86)
- Improving Single Domain-Generalized Object Detection: A Focus on Diversification and Alignment (CVPR 2024 · similarity 0.85)
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.