Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Zero-Shot Learning from Adversarial Feature Residual to Compact Visual Feature
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 feature residuals and add them to predicted visual prototypes instead of generating visual features directly (ablation) | AFRNet · CUB SCS, no feature selection | Top-1 accuracy: 41.6 -> 48.7 | Helped |
| Generate feature residuals and add them to predicted visual prototypes instead of generating visual features directly (ablation) | AFRNet · CUB SCE, no feature selection | Top-1 accuracy: 9.1 -> 18.6 | Helped |
| Generate feature residuals and add them to predicted visual prototypes instead of generating visual features directly (ablation) | AFRNet · NAB SCS, no feature selection | Top-1 accuracy: 37.1 -> 41.5 | Helped |
| Generate feature residuals and add them to predicted visual prototypes instead of generating visual features directly (ablation) | AFRNet · NAB SCE, no feature selection | Top-1 accuracy: 5.7 -> 12.7 | Helped |
| Add feature selection based on prototype prediction error (ablation) | AFRNet · CUB and NAB, SCS/SCE; 1NN and AFRNet classifiers | Top-1 accuracy: 44.3 -> 48.7; 48.7 -> 50.3; 16.3 -> 18.2; 18.6 -> 20.5; 35.2 -> 38.2; 41.5 -> 42.8; 9.4 -> 9.8; 12.7 -> 12.8 | Helped |
| Use AFRNet rather than the strongest listed baseline | AFRNet · APY, AWA1, AWA2, SUN (PS-split, conventional ZSL) | Top-1 accuracy: 44.0 -> 56.2; 72.0 -> 76.4; 65.2 -> 75.1; 62.8 -> 64.0 | Helped |
| Use AFRNet rather than the strongest listed baseline | AFRNet · CUB and NAB, SCS/SCE (conventional ZSL) | Top-1 accuracy: 43.7 -> 50.3; 10.9 -> 20.5; 39.4 -> 42.8; 9.7 -> 12.8 | Helped |
| Use AFRNet rather than the strongest listed baseline | AFRNet · APY, AWA1, AWA2 (PS-split, generalized ZSL) | Harmonic mean: 45.7 -> 58.9; 62.3 -> 68.8; 46.9 -> 70.1 | Helped |
| Use AFRNet rather than SABR | AFRNet · SUN (PS-split, generalized ZSL) | Harmonic mean: 41.5 vs 41.5 | No clear effect |
Closest papers by meaning
- Zero-Shot Visual Recognition Using Semantics-Preserving Adversarial Embedding Networks (CVPR 2018 · similarity 0.90)
- Generative Dual Adversarial Network for Generalized Zero-Shot Learning (CVPR 2019 · similarity 0.90)
- Feature Generating Networks for Zero-Shot Learning (CVPR 2018 · similarity 0.90)
- ZeroDiff: Solidified Visual-semantic Correlation in Zero-Shot Learning (ICLR 2025 · similarity 0.88)
- Progressive Semantic-Guided Vision Transformer for Zero-Shot Learning (CVPR 2024 · similarity 0.88)
- Data-Free Generalized Zero-Shot Learning (AAAI 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.