Research map / Detection and segmentation / CNN features
Dense Distinct Query 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 |
|---|---|---|---|
| add pyramid shuffle for cross-level query interaction (ablation) | FCOS* · MS-COCO val, 1x schedule | AP: 36.5 -> 37.6 | Helped |
| add distinct queries selection (ablation) | FCOS* with pyramid shuffle · MS-COCO val, 1x schedule | AP: 37.6 -> 40.6 | Helped |
| add auxiliary loss for dense queries (ablation) | DDQ FCN · MS-COCO val, 1x schedule | AP: 40.6 -> 41.5 | Helped |
| increase query count to 7000 (ablation) | Sparse R-CNN · MS-COCO val, 1x schedule | AP: 39.4 -> 40.6 | Helped |
| apply distinct queries selection at each stage (ablation) | Sparse R-CNN with 7000 queries · MS-COCO val, 1x schedule | AP: 40.6 -> 43.1 | Helped |
| replace the first four refinement stages with DDQ FCN (ablation) | Sparse R-CNN with DQS · MS-COCO val, 1x schedule | AP: 43.1 -> 44.6 | Helped |
| initialize dense query content with feature embeddings rather than mapped coordinates (ablation) | Deformable DETR* · MS-COCO val, DETR augmentation, 12 epochs | AP: 46.7 -> 48.5 | Helped |
| add auxiliary loss for dense decoder queries (ablation) | DDQ DETR with dense feature content · MS-COCO val, DETR augmentation, 12 epochs | AP: 48.5 -> 50.0 | Helped |
| apply DQS before each refinement stage (ablation) | DDQ DETR · MS-COCO val, DETR augmentation, 12 epochs | AP: 50.0 -> 50.7 | Helped |
| add P2 feature and 100 CDN queries (ablation) | DDQ DETR · MS-COCO val, DETR augmentation, 12 epochs | AP: 50.7 -> 52.1 | Helped |
| remove distinct queries selection (ablation) | DDQ FCN · MS-COCO val | AP: 41.5 -> 39.5* | Hurt |
| remove distinct queries selection (ablation) | DDQ R-CNN · MS-COCO val | AP: 44.6 -> 42.7* | Hurt |
Closest papers by meaning
- Progressive End-to-End Object Detection in Crowded Scenes (CVPR 2022 · similarity 0.89)
- CQ-DINO: Mitigating Gradient Dilution via Category Queries for Vast Vocabulary Object Detection (NeurIPS 2025 · similarity 0.85)
- Sparse Semi-DETR: Sparse Learnable Queries for Semi-Supervised Object Detection (CVPR 2024 · similarity 0.85)
- BEVNeXt: Reviving Dense BEV Frameworks for 3D Object Detection (CVPR 2024 · similarity 0.85)
- Order Matters: Competition-Guided Query Ordering for RNN-Based Object Detection (NeurIPS 2026 · similarity 0.84)
- Reg-PTQ: Regression-specialized Post-training Quantization for Fully Quantized Object Detector (CVPR 2024 · similarity 0.84)
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.