Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Transductive Unbiased Embedding for Zero-Shot Learning
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 |
|---|---|---|---|
| Add the unlabeled-target bias loss to source-only training (ablation) | QFSL · CUB, AwA2, and SUN; conventional SS and PS | MCA_t: 58.5 -> 69.7; 58.8 -> 72.1; 58.9 -> 61.7; 56.2 -> 58.3; 72.6 -> 84.8; 63.5 -> 79.7 | Helped |
| Use unlabeled target images in QFSL training | QFSL · AwA2 generalized | H: +16.7 | Helped |
| Use unlabeled target images in QFSL training | QFSL · CUB generalized | H: +24.5 | Helped |
| Use unlabeled target images in QFSL training | QFSL · SUN generalized | H: +12.2 | Helped |
| Optimize the visual embedding subnet instead of keeping it fixed (ablation) | QFSL · CUB and AwA2 generalized | MCA | Helped |
| Optimize the visual embedding subnet instead of keeping it fixed (ablation) | QFSL · SUN generalized | MCA | Hurt |
| Increase the bias-loss weight above 2 (ablation) | QFSL · AwA2, CUB, and SUN generalized | MCA_t | Hurt |
| Increase the bias-loss weight (ablation) | QFSL · AwA2, CUB, and SUN generalized | MCA_s | Hurt |
| Set the bias-loss weight below 0.5 (ablation) | QFSL · AwA2, CUB, and SUN generalized | MCA_t | Hurt |
| Increase the number of source classes (ablation) | QFSL · SUN generalized | MCA_t | Helped |
Closest papers by meaning
- Semantically Aligned Bias Reducing Zero Shot Learning (CVPR 2019 · similarity 0.91)
- Attentive Region Embedding Network for Zero-Shot Learning (CVPR 2019 · similarity 0.89)
- Improving Generalized Zero-Shot Learning by Exploring the Diverse Semantics from External Class Names (CVPR 2024 · similarity 0.88)
- ZeroDiff: Solidified Visual-semantic Correlation in Zero-Shot Learning (ICLR 2025 · similarity 0.87)
- Data-Free Generalized Zero-Shot Learning (AAAI 2024 · similarity 0.87)
- Epsilon: Exploring Comprehensive Visual-Semantic Projection for Multi-Label Zero-Shot Learning (AAAI 2025 · similarity 0.87)
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.