Research map / Detection and segmentation / CNN features
Q-DETR: An Efficient Low-Bit Quantized Detection Transformer
Sheng Xu, Yanjing Li, Mingbao Lin, Peng Gao, Guodong Guo, Jinhu Lü, Baochang 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 |
|---|---|---|---|
| Add distribution rectification distillation to 4-bit DETR-R50 | DETR-R50 · COCO val2017 | AP: 34.1 -> 39.4 | Improved |
| Add distribution rectification distillation to 3-bit DETR-R50 | DETR-R50 · COCO val2017 | AP: 32.3 -> 36.1 | Improved |
| Add distribution rectification distillation to 2-bit DETR-R50 | DETR-R50 · COCO val2017 | AP: 26.6 -> 31.4 | Improved |
| Add distribution rectification distillation to 4-bit SMCA-DETR-R50 | SMCA-DETR-R50 · COCO val2017 | AP: 35.0 -> 38.3 | Improved |
| Add distribution alignment to 2-bit DETR-R50 (ablation) | DETR-R50 · VOC test2007 | AP50: 69.7 -> 71.6 | Improved |
| Add foreground-aware query matching to 2-bit DETR-R50 (ablation) | DETR-R50 · VOC test2007 | AP50: 69.7 -> 74.9 | Improved |
| Add both distribution alignment and foreground-aware query matching to 2-bit DETR-R50 (ablation) | DETR-R50 · VOC test2007 | AP50: 69.7 -> 76.4 | Improved |
| Add distribution alignment to 4-bit DETR-R50 (ablation) | DETR-R50 · VOC test2007 | AP50: 78.0 -> 78.8 | Improved |
| Add foreground-aware query matching to 4-bit DETR-R50 (ablation) | DETR-R50 · VOC test2007 | AP50: 78.0 -> 81.5 | Improved |
| Add both distribution alignment and foreground-aware query matching to 4-bit DETR-R50 (ablation) | DETR-R50 · VOC test2007 | AP50: 78.0 -> 82.7 | Improved |
| Use full imitation with query-selection threshold τ=0 instead of no distillation at τ=1 (ablation) | 4-bit DETR-R50 + distribution alignment · VOC test2007 | AP50 | Hurt |
| Quantize the MHA decoder to low bits (ablation) | DETR-R50 · VOC | AP: 2.1% | Hurt |
Closest papers by meaning
- AQ-DETR: Low-Bit Quantized Detection Transformer with Auxiliary Queries (AAAI 2024 · similarity 0.94)
- KD-DETR: Knowledge Distillation for Detection Transformer with Consistent Distillation Points Sampling (CVPR 2024 · similarity 0.88)
- Salience DETR: Enhancing Detection Transformer with Hierarchical Salience Filtering Refinement (CVPR 2024 · similarity 0.87)
- DETRs Beat YOLOs on Real-time Object Detection (CVPR 2024 · similarity 0.87)
- Reg-PTQ: Regression-specialized Post-training Quantization for Fully Quantized Object Detector (CVPR 2024 · similarity 0.87)
- MI-DETR: An Object Detection Model with Multi-time Inquiries Mechanism (CVPR 2025 · 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.
