Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Zero-Shot Learning via Class-Conditioned Deep Generative Models
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 |
|---|---|---|---|
| Use class-conditioned latent distributions and a VAE reconstruction objective instead of point-based class representations | VZSL · Inductive ZSL: AwA, CUB-200, SUN, and ImageNet | Top-1 accuracy (%) on AwA/CUB-200/SUN; Top-5 accuracy (%) on ImageNet: 80.83 -> 85.28 (AwA); 56.60 -> 57.42 (CUB-200); 86.50 -> 86.75 (SUN); 16.8 -> 23.08 (ImageNet) | Helped |
| Add the VAE reconstruction term (ablation) | VZSL · Inductive ZSL: AwA, CUB-200, SUN, and ImageNet | Top-1 accuracy (%) on AwA/CUB-200/SUN; Top-5 accuracy (%) on ImageNet: 84.45± 0.74 -> 85.28± 0.76; 55.37± 0.59 -> 57.42± 0.63; 85.75± 1.93 -> 86.75± 2.02; 22.88 -> 23.08 | Helped |
| Add the VAE reconstruction term (ablation) | VZSL · Transductive ZSL: AwA, CUB-200, and SUN | Top-1 accuracy (%): 93.49± 0.54 -> 94.80± 0.17; 59.69± 1.22 -> 66.45± 0.88; 86.37± 1.88 -> 87.75± 1.43 | Helped |
| Add the unlabeled-data KL sharpening regularizer (ablation) | VZSL transductive objective · AwA and CUB-200, transductive ZSL | Top-1 accuracy (%): 87.59± 0.21 -> 94.80± 0.17; 61.44 ± 0.98 -> 66.45± 0.88 | Helped |
| Add the unlabeled-data KL sharpening regularizer (ablation) | VZSL transductive objective · SUN, transductive ZSL | Top-1 accuracy (%): 86.66± 1.67 -> 87.75± 1.43 | Helped |
| Provide labeled examples from unseen classes | VZSL few-shot learning · Few-shot learning | Accuracy | Helped |
Closest papers by meaning
- Zero-shot Learning via Simultaneous Generating and Learning (NeurIPS 2019 · similarity 0.95)
- GenZSL: Generative Zero-Shot Learning Via Inductive Variational Autoencoder (ICML 2025 · similarity 0.91)
- Improving Generalized Zero-Shot Learning by Exploring the Diverse Semantics from External Class Names (CVPR 2024 · similarity 0.91)
- Data-Free Generalized Zero-Shot Learning (AAAI 2024 · similarity 0.90)
- Class Distribution Shifts in Zero-Shot Learning: Learning Robust Representations (NeurIPS 2024 · similarity 0.90)
- Data Distribution Distilled Generative Model for Generalized Zero-Shot Recognition (AAAI 2024 · 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.