Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Leveraging Seen and Unseen Semantic Relationships for Generative Zero-Shot Learning
Maunil R Vyas, Hemanth Venkateswara, Sethuraman Panchanathan
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 |
|---|---|---|---|
| add SR-Loss semantic relationship regularization (vs F-GAN) on unseen-class confidence | LsrGAN · AWA (GZSL) | average softmax confidence on unseen classes: 0.29 -> 0.69 | Improved |
| add SR-Loss (vs F-GAN) on unseen-class confidence | LsrGAN · CUB (GZSL, attributes) | average softmax confidence on unseen classes: 0.33 -> 0.60 | Improved |
| add SR-Loss (vs F-GAN) on unseen-class confidence | LsrGAN · SUN (GZSL) | average softmax confidence on unseen classes: 0.32 -> 0.65 | Improved |
| add SR-Loss (vs F-GAN) on seen-class confidence | LsrGAN · AWA (GZSL) | average softmax confidence on seen classes: 0.86 -> 0.83 | Hurt |
| add SR-Loss (vs F-GAN) on seen-class confidence | LsrGAN · SUN (GZSL) | average softmax confidence on seen classes: 0.35 -> 0.39 | Improved |
| add SR-Loss and related components versus LisGAN, ZSL | LsrGAN · AWA (attributes, ZSL) | Top-1 accuracy: 70.6 -> 66.4 | Hurt |
| LsrGAN versus LisGAN on GZSL harmonic mean | LsrGAN · CUB (attributes, GZSL) | harmonic mean: 51.6 -> 53.0 | Improved |
| LsrGAN versus GAZSL on Wikipedia CUB Hard ZSL | LsrGAN · CUB Hard (Wikipedia, ZSL) | Top-1 accuracy: 10.3 -> 14.2 | Improved |
| LsrGAN versus GAZSL on Wikipedia CUB Easy GZSL | LsrGAN · CUB Easy (Wikipedia, GZSL) | seen-unseen AUC: 35.4 -> 39.5 | Improved |
| LsrGAN versus GAZSL on NAB Hard GZSL | LsrGAN · NAB Hard (Wikipedia, GZSL) | seen-unseen AUC: 5.8 -> 6.4 | Improved |
| add visual pivot regularizer and SR-Loss, and denoiser (ablation S1-S4, values only in Fig. 4a) (ablation) | LsrGAN · CUB (ZSL ablation) | Top-1 accuracy | Improved |
| vary soft margin epsilon of SR-Loss (ablation) | LsrGAN · SR-Loss parameter sensitivity | accuracy | Mixed |
Closest papers by meaning
- Adaptive and Generative Zero-Shot Learning (ICLR 2021 · similarity 0.94)
- SimZSL: Zero-Shot Learning Beyond a Pre-defined Semantic Embedding Space (IJCV 2025 · similarity 0.93)
- Improving Generalized Zero-Shot Learning by Exploring the Diverse Semantics from External Class Names (CVPR 2024 · similarity 0.92)
- ZeroDiff: Solidified Visual-semantic Correlation in Zero-Shot Learning (ICLR 2025 · similarity 0.92)
- Progressive Semantic-Guided Vision Transformer for Zero-Shot Learning (CVPR 2024 · similarity 0.91)
- ZeroMamba: Exploring Visual State Space Model for Zero-Shot Learning (AAAI 2025 · similarity 0.91)
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.
