Research map / Detection and segmentation / Open-set and few-shot detection
Transductive Learning for Zero-Shot Object Detection
Shafin Rahman, Salman Khan, Nick Barnes
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 transductive pseudo-label learning | FL-65 inductive RetinaNet · MSCOCO ZSD (65/15 split) | mAP: 10.80 -> 14.57 | Improved |
| add transductive pseudo-label learning | FL-65 inductive RetinaNet · MSCOCO GZSD (65/15 split) | harmonic mean mAP: 16.77 -> 18.89 | Improved |
| add unseen pseudo-labeling and unseen maximization to seen-only pseudo-labeling (ablation) | Transductive baseline · MSCOCO ZSD (65/15 split) | mAP: 12.40 -> 14.57 | Improved |
| add Ld(u) to Ld(s)+L'd(u) (ablation) | Dynamic pseudo-labeling · MSCOCO ZSD (65/15 split) | mAP: 12.60 -> 14.57 | Improved |
| add L'd(u) to Ld(s)+Ld(u) (ablation) | Dynamic pseudo-labeling · MSCOCO ZSD (65/15 split) | mAP: 13.22 -> 14.57 | Improved |
| add Ld(s) to Ld(u)+L'd(u) (ablation) | Dynamic pseudo-labeling · MSCOCO GZSD (65/15 split) | harmonic mean mAP: 15.96 -> 18.89 | Improved |
| use Ld(u) alone as dynamic component (ablation) | Transductive model · MSCOCO ZSD (65/15 split) | mAP: 12.40 -> 10.97 | Hurt |
| use L'd(u) alone as dynamic component (ablation) | Transductive model · MSCOCO ZSD (65/15 split) | mAP: 12.40 -> 9.47 | Hurt |
| add seen-only dynamic pseudo-labeling to inductive model (ablation) | FL-65 · MSCOCO ZSD (65/15 split) | mAP: 10.8 -> 12.4 | Improved |
| freeze rest of network and tune only classification subnet (ablation) | Transductive ZSD · MSCOCO ZSD | mAP | Improved |
| use GloVe vectors instead of word2vec | Transductive ZSD · MSCOCO ZSD (65/15 split) | mAP: 10.36 -> 12.87 | Improved |
| add transductive learning | Zero-shot detector · PASCAL VOC ZSD (16/4 split) | average unseen-class mAP: 62.1 -> 66.6 | Improved |
Closest papers by meaning
- Zero-Shot Object Detection (ECCV 2018 · similarity 0.92)
- Meta-ZSDETR: Zero-shot DETR with Meta-learning (ICCV 2023 · similarity 0.92)
- Semantics-Guided Contrastive Network for Zero-Shot Object Detection (TPAMI 2022 · similarity 0.91)
- Transductive Zero-Shot Learning with Visual Structure Constraint (NeurIPS 2019 · similarity 0.89)
- An Empirical Study and Analysis of Generalized Zero-Shot Learning for Object Recognition in the Wild (ECCV 2016 · similarity 0.89)
- Transductive Zero-Shot and Few-Shot CLIP (CVPR 2024 · similarity 0.88)
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.
