Research map / Detection and segmentation / Open-set and few-shot detection
Open-Vocabulary DETR with Conditional Matching
Yuhang Zang, Wei Li, Kaiyang Zhou, Chen Huang, Chen Change Loy
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 classifier with CLIP text embeddings (ablation) | Deformable DETR · OV-LVIS | mask AP novel: 0.0 -> 9.5 | Improved |
| add unlabeled object proposals under closed-set matching (ablation) | Deformable DETR + CLIP text · OV-LVIS | mask AP novel: 9.5 -> 6.3 | Hurt |
| add conditional binary matching on top of proposals (ablation) | Deformable DETR + proposals · OV-LVIS | mask AP novel: 9.5 -> 17.4 | Improved |
| add embedding reconstruction loss L_embed (ablation) | OV-DETR · OV-LVIS | mask AP novel: 14.4 -> 17.4 | Improved |
| copy object queries R=3 times instead of 1 (N=100) (ablation) | OV-DETR · OV-LVIS | mask AP novel: 10.6 -> 13.6 | Improved |
| increase copies from R=3 to R=9 (N=100) (ablation) | OV-DETR · OV-LVIS | mask AP novel: 13.6 -> 11.9 | Hurt |
| increase queries N from 100 to 300 (R=3) (ablation) | OV-DETR · OV-LVIS | mask AP novel: 13.6 -> 17.4 | Improved |
| copy object queries R=3 instead of 1 (N=300) (ablation) | OV-DETR · OV-LVIS | mask AP novel: 12.3 -> 17.4 | Improved |
| optimize inference by forwarding multiple conditional queries in parallel (ablation) | OV-DETR · LVIS | inference time (s/iter): 23.84 -> 9.57 | Improved |
| OV-DETR versus ViLD on OV-COCO | OV-DETR · OV-COCO | AP50 novel: 27.6 -> 29.4 | Improved |
| OV-DETR versus ViLD on transfer to PASCAL VOC | OV-DETR · PASCAL VOC 2007 test | AP50: 72.2 -> 76.1 | Improved |
| Deformable DETR instead of Mask R-CNN (base-only training) (ablation) | base detector · OV-LVIS | mask AP: 22.5 -> 22.4 | No clear effect |
Closest papers by meaning
- Open-Det: An Efficient Learning Framework for Open-Ended Detection (ICML 2025 · similarity 0.92)
- DetCLIPv3: Towards Versatile Generative Open-vocabulary Object Detection (CVPR 2024 · similarity 0.91)
- ProxyDet: Synthesizing Proxy Novel Classes via Classwise Mixup for Open-Vocabulary Object Detection (AAAI 2024 · similarity 0.91)
- OVMR: Open-Vocabulary Recognition with Multi-Modal References (CVPR 2024 · similarity 0.90)
- Exploring Region-Word Alignment in Built-in Detector for Open-Vocabulary Object Detection (CVPR 2024 · similarity 0.90)
- OV-DQUO: Open-Vocabulary DETR with Denoising Text Query Training and Open-World Unknown Objects Supervision (AAAI 2025 · similarity 0.90)
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.
