Research map / Detection and segmentation / CNN features
Accelerating DETR Convergence via Semantic-Aligned Matching
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 semantic-aligned matching with query resampling (ablation) | DETR · COCO val 2017, 12 epochs | AP: 22.3 -> 25.2 | Helped |
| Replace average-pooling with max-pooling resampling (ablation) | SAM-DETR · COCO val 2017, 12 epochs | AP: 25.2 -> 27.0 | Helped |
| Use one salient point instead of max-pooling (ablation) | SAM-DETR · COCO val 2017, 12 epochs | AP: 27.0 -> 28.6 | Helped |
| Add reweighting by previous query embeddings (ablation) | SAM-DETR with one salient point · COCO val 2017, 12 epochs | AP: 28.6 -> 30.3 | Helped |
| Use eight salient points instead of one (ablation) | SAM-DETR · COCO val 2017, 12 epochs | AP: 28.6 -> 32.0 | Helped |
| Add reweighting by previous query embeddings (ablation) | SAM-DETR with eight salient points · COCO val 2017, 12 epochs | AP: 32.0 -> 33.1 | Helped |
| Constrain salient-point search to reference boxes rather than the full image (ablation) | SAM-DETR · COCO val 2017, 12 epochs | AP: 30.0 -> 33.1 | Helped |
| Add semantic-aligned matching | DETR-R50 · COCO 2017 val, R50, 12 epochs | AP: +10.8% AP | Helped |
| Integrate SAM-DETR with SMCA | SAM-DETR-R50 · COCO 2017 val, R50, 12 epochs | AP: +2.9% AP | Helped |
| Integrate SAM-DETR with SMCA | SAM-DETR-R50 · COCO 2017 val, R50, 12 epochs | AP: 33.1 -> 36.0 | Helped |
Closest papers by meaning
- Sparse Semi-DETR: Sparse Learnable Queries for Semi-Supervised Object Detection (CVPR 2024 · similarity 0.88)
- Not All Tokens Matter All The Time: Dynamic Token Aggregation Towards Efficient Detection Transformers (ICML 2025 · similarity 0.88)
- Mean Teacher DETR with Masked Feature Alignment: A Robust Domain Adaptive Detection Transformer Framework (AAAI 2024 · similarity 0.88)
- Semantic-Aware Transformation-Invariant RoI Align (AAAI 2024 · similarity 0.88)
- Open-Det: An Efficient Learning Framework for Open-Ended Detection (ICML 2025 · similarity 0.87)
- MS-DETR: Efficient DETR Training with Mixed Supervision (CVPR 2024 · similarity 0.87)
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.