Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Vocabulary-Informed Zero-Shot and Open-Set Learning
Yanwei Fu, Xiaomei Wang, Hanze Dong, Yu–Gang Jiang, Meng Wang, Xiangyang Xue, Leonid Sigal
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 |
|---|---|---|---|
| learn constraints from the open vocabulary instead of known classes only (ablation) | WMM-Voc · AwA zero-shot | accuracy (%): 84.51 -> 90.79 | Improved |
| learn constraints from the open vocabulary instead of known classes only (ablation) | Deep WMM-Voc · AwA zero-shot | accuracy (%): 83.85 -> 90.65 | Improved |
| add density weighting (WMM-Voc over MM-Voc) (ablation) | MM-Voc · AwA zero-shot, 200 instances, 1000-dim | top-1 accuracy (%): 83.84 -> 89.09 | Improved |
| add density weighting (WMM-Voc over MM-Voc) (ablation) | MM-Voc · AwA G-ZSL, 1000-dim | top-1 accuracy (%): 8.54 -> 27.98 | Improved |
| add density weighting (WMM-Voc over MM-Voc) (ablation) | MM-Voc · AwA open-set | false positive rate: 0.12 -> 0.05 | Improved |
| fine-tune CNN features (Deep WMM-Voc over WMM-Voc) (ablation) | WMM-Voc · AwA zero-shot, 100-dim | top-1 accuracy (%): 82.78 -> 84.87 | Improved |
| fine-tune CNN features (Deep WMM-Voc over WMM-Voc) (ablation) | WMM-Voc · AwA zero-shot, 1000-dim | top-1 accuracy (%): 89.09 -> 88.07 | Hurt |
| add density weighting (WMM-Voc over MM-Voc) (ablation) | MM-Voc · ImageNet supervised | top-1 accuracy (%): 37.1 -> 35.95 | Hurt |
| fine-tune features (Deep WMM-Voc over WMM-Voc) (ablation) | WMM-Voc · ImageNet supervised | top-1 accuracy (%): 35.95 -> 38.92 | Improved |
| use vocabulary-informed learning instead of ESZSL | ESZSL · AwA G-ZSL, 1000-dim | top-1 accuracy (%): 0 -> 34.77 | Improved |
| add vocabulary constraints with more training instances | Voc methods · AwA supervised | accuracy | Mixed |
| use hybrid SGD/L-BFGS solver instead of L-BFGS | Voc · large datasets | training time: 20~50% | Improved |
Closest papers by meaning
- Semi-Supervised Vocabulary-Informed Learning (CVPR 2016 · similarity 0.97)
- Synthesized Classifiers for Zero-Shot Learning (CVPR 2016 · similarity 0.92)
- SimZSL: Zero-Shot Learning Beyond a Pre-defined Semantic Embedding Space (IJCV 2025 · similarity 0.92)
- TPR: Topology-Preserving Reservoirs for Generalized Zero-Shot Learning (NeurIPS 2024 · similarity 0.90)
- S3A: Towards Realistic Zero-Shot Classification via Self Structural Semantic Alignment (AAAI 2024 · similarity 0.89)
- Label Propagation for Zero-shot Classification with Vision-Language Models (CVPR 2024 · 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.
