Research map / Detection and segmentation / CNN features
Towards Data-Efficient Detection Transformers
Wen Wang, Jing Zhang, Yang Cao, Yongliang Shen, Dacheng Tao
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 sparse feature sampling (RoIAlign) (ablation) | CondDETR · Cityscapes | AP: 12.1 -> 20.4 | Improved |
| add multi-scale feature fusion (ablation) | DE-CondDETR · Cityscapes | AP: 20.4 -> 26.8 | Improved |
| add label augmentation (ablation) | DE-CondDETR · Cityscapes | AP: 26.8 -> 29.5 | Improved |
| add label augmentation alone (ablation) | CondDETR · Cityscapes | AP: 12.1 -> 14.7 | Improved |
| add RoIAlign alone (no box refinement) (ablation) | CondDETR · Cityscapes | AP: 12.1 -> 16.3 | Improved |
| add box refinement alone (ablation) | CondDETR · Cityscapes | AP: 12.1 -> 12.1 | No clear effect |
| add box refinement on top of RoIAlign (ablation) | CondDETR · Cityscapes | AP: 16.3 -> 20.4 | Improved |
| increase RoIAlign resolution from 1 to 4 (ablation) | DE-CondDETR · Cityscapes | AP: 14.8 -> 20.4 | Improved |
| increase RoIAlign resolution from 4 to 7 (ablation) | DE-CondDETR · Cityscapes | AP: 20.4 -> 20.7 | No clear effect |
| add a 4th (64x) feature level (ablation) | DE-CondDETR · Cityscapes | AP: 26.8 -> 26.3 | Hurt |
| apply label augmentation with only 100 queries (ablation) | DE-DETR · Cityscapes | AP: 21.7 -> 20.6 | Hurt |
| increase queries to 300 and add label augmentation (ablation) | DE-DETR · Cityscapes | AP: 22.4 -> 24.5 | Improved |
Closest papers by meaning
- End-to-End Object Detection with Transformers (ECCV 2020 · similarity 0.89)
- Efficient Decoder-Free Object Detection with Transformers (ECCV 2022 · similarity 0.88)
- CF-DETR: Coarse-to-Fine Transformers for End-to-End Object Detection (AAAI 2022 · similarity 0.87)
- DESTR: Object Detection With Split Transformer (CVPR 2022 · similarity 0.87)
- KD-DETR: Knowledge Distillation for Detection Transformer with Consistent Distillation Points Sampling (CVPR 2024 · similarity 0.86)
- Context Enhanced Transformer for Single Image Object Detection in Video Data (AAAI 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.
