Research map / Detection and segmentation / Open-set and few-shot detection
Meta-ZSDETR: Zero-shot DETR with Meta-learning
Lu Zhang, Chenbo Zhang, Jiajia Zhao, Jihong Guan, Shuigeng Zhou
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 classification loss to filter class-specific boxes (ablation) | Meta-ZSDETR · MS COCO 65/15 GZSD unseen | mAP: 14.5 -> 20.6 | Improved |
| add contrastive-reconstruction loss (ablation) | Meta-ZSDETR · MS COCO 65/15 GZSD unseen | mAP: 20.6 -> 21.7 | Improved |
| train regression head only on positive predictions (ablation) | Meta-ZSDETR · MS COCO 65/15 GZSD unseen | mAP: 16.5 -> 21.7 | Improved |
| train classification head on all predictions including background (ablation) | Meta-ZSDETR · MS COCO 65/15 GZSD unseen | mAP: 17.9 -> 21.7 | Improved |
| exclude background predictions from contrastive head (ablation) | Meta-ZSDETR · MS COCO 65/15 GZSD unseen | mAP: 21.3 -> 21.7 | Improved |
| contrastive head with positives only versus none (ablation) | Meta-ZSDETR · MS COCO 65/15 GZSD unseen | mAP: 20.6 -> 21.0 | Improved |
| use DETR with class-specific queries instead of Faster R-CNN proposals | zero-shot detection · PASCAL VOC 16/4 ZSD | mAP: 65.7 -> 70.3 | Improved |
| increase number of queries N (ablation) | Meta-ZSDETR · MS COCO unseen mAP | mAP | Improved |
| use positive rate 0.5 with N=900 (ablation) | Meta-ZSDETR · MS COCO unseen mAP | mAP | Improved |
Closest papers by meaning
- Zero-Shot Object Detection (ECCV 2018 · similarity 0.93)
- Transductive Learning for Zero-Shot Object Detection (ICCV 2019 · similarity 0.92)
- Semantics-Guided Contrastive Network for Zero-Shot Object Detection (TPAMI 2022 · similarity 0.91)
- Meta-DETR: Image-Level Few-Shot Detection with Inter-Class Correlation Exploitation (TPAMI 2022 · similarity 0.91)
- Meta-Learning to Detect Rare Objects (ICCV 2019 · similarity 0.90)
- Rethinking Zero-Shot Learning: A Conditional Visual Classification Perspective (ICCV 2019 · similarity 0.89)
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.
