Research map / Detection and segmentation / CNN features
DAB-DETR: Dynamic Anchor Boxes are Better Queries for DETR
Shilong Liu, Feng Li, Hao Zhang, Xiao Yang, Xianbiao Qi, Hang Su, Jun Zhu, Lei Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 learned positional queries with dynamic anchor-box queries | DETR · COCO detection, ResNet-50-DC5 | AP | Improved |
| Add 3 pattern embeddings | DAB-DETR · COCO 2017 val2017, ResNet-50-DC5 | AP: 44.5 -> 45.7 | Improved |
| Add iterative anchor update (ablation) | DAB-DETR · COCO 2017 val2017, ResNet-50-DC5 | AP: 44.0 -> 45.7 | Improved |
| Add width- and height-modulated attention (ablation) | DAB-DETR · COCO 2017 val2017, ResNet-50-DC5 | AP: 45.0 -> 45.7 | Improved |
| Use 4D anchor boxes instead of 2D anchor points (ablation) | DAB-DETR · COCO 2017 val2017, ResNet-50-DC5 | AP: 44.0 -> 45.0 | Improved |
| Tune positional-encoding temperature (ablation) | DAB-DETR · COCO, ResNet-50, no pattern embeddings or modulated attention | AP: 39.5 -> 40.1 | Improved |
| Increase decoder depth from 2 to 6 layers (ablation) | DAB-DETR · COCO 2017 val2017, ResNet-50-DC5 | AP: 40.2 -> 45.7 | Improved |
| Fix initial anchor centers x,y to random values (ablation) | DAB-DETR · COCO 2017 val2017, tested backbones | AP: +0.3, +0.2, +0.1, +0.7, +0.1 | Improved |
| Replace Deformable DETR query formulation with dynamic anchor-box queries | Deformable DETR · COCO 2017 val2017, ResNet-50 multi-scale | AP: 46.3 to 46.8 | Improved |
| Increase temperature to flatten positional attention (ablation) | DAB-DETR · COCO, object sizes | AP by object size: AP$_{S}$ 19.3 vs. 18.9; AP$_{L}$ 58.0 vs. 58.9 | Mixed |
Closest papers by meaning
- DAC-DETR: Divide the Attention Layers and Conquer (NeurIPS 2023 · similarity 0.90)
- Anchor DETR: Query Design for Transformer-Based Detector (AAAI 2022 · similarity 0.89)
- Salience DETR: Enhancing Detection Transformer with Hierarchical Salience Filtering Refinement (CVPR 2024 · similarity 0.88)
- AQ-DETR: Low-Bit Quantized Detection Transformer with Auxiliary Queries (AAAI 2024 · similarity 0.88)
- EASE-DETR: Easing the Competition among Object Queries (CVPR 2024 · similarity 0.87)
- Order Matters: Competition-Guided Query Ordering for RNN-Based Object Detection (NeurIPS 2026 · 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.
