Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
A Generative Adversarial Approach for Zero-Shot Learning From Noisy Texts
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 |
|---|---|---|---|
| generate unseen-class features with GAZSL instead of using ZSLPP | zero-shot recognition · CUB SCS-split | top-1 accuracy (%): 37.2 -> 43.7 | Helped |
| generate unseen-class features with GAZSL instead of using ZSLPP | zero-shot recognition · CUB SCE-split | top-1 accuracy (%): 9.7 -> 10.3 | Helped |
| generate unseen-class features with GAZSL instead of using ZSLPP | zero-shot recognition · NAB SCS-split | top-1 accuracy (%): 30.3 -> 35.6 | Helped |
| generate unseen-class features with GAZSL instead of using ZSLPP | zero-shot recognition · NAB SCE-split | top-1 accuracy (%): 8.1 -> 8.6 | Helped |
| use GAZSL instead of ZSLPP | zero-shot image retrieval · CUB SCS-split, retrieval 25% | mAP (%): 42.3 -> 49.7 | Helped |
| use GAZSL instead of ZSLPP | zero-shot image retrieval · NAB SCS-split, retrieval 100% | mAP (%): 31.3 -> 31.0 | Hurt |
| add visual pivot regularization (ablation) | GAZSL · CUB SCS-split | zero-shot classification accuracy (%): 22.83 -> 43.74 | Helped |
| add visual pivot regularization (ablation) | GAZSL · NAB SCS-split | zero-shot classification accuracy (%): 24.22 -> 35.58 | Helped |
| add the text FC layer (ablation) | GAZSL · CUB SCS-split | zero-shot classification accuracy (%): 40.85 -> 43.74 | Helped |
| add the text FC layer (ablation) | GAZSL · NAB SCS-split | zero-shot classification accuracy (%): 32.94 -> 35.58 | Helped |
| remove GAN adversarial training while retaining visual pivot loss (ablation) | GAZSL · CUB SCS-split | zero-shot classification accuracy (%): 43.74 -> 28.52 | Hurt |
| remove GAN adversarial training while retaining visual pivot loss (ablation) | GAZSL · NAB SCS-split | zero-shot classification accuracy (%): 35.58 -> 25.75 | Hurt |
Closest papers by meaning
- Feature Generating Networks for Zero-Shot Learning (CVPR 2018 · similarity 0.92)
- Leveraging the Invariant Side of Generative Zero-Shot Learning (CVPR 2019 · similarity 0.91)
- On the Zero-shot Adversarial Robustness of Vision-Language Models: A Truly Zero-shot and Training-free Approach (CVPR 2025 · similarity 0.87)
- GenZSL: Generative Zero-Shot Learning Via Inductive Variational Autoencoder (ICML 2025 · similarity 0.86)
- Data-Free Generalized Zero-Shot Learning (AAAI 2024 · similarity 0.86)
- ZeroDiff: Solidified Visual-semantic Correlation in Zero-Shot Learning (ICLR 2025 · similarity 0.86)
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.