Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
VGSE: Visually-Grounded Semantic Embeddings for Zero-Shot Learning
Check 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 |
|---|---|---|---|
| Replace w2v class embeddings with VGSE-SMO embeddings | SJE · AWA2 | ZSL Top-1 accuracy: 53.7 -> 62.4 | Helped |
| Replace w2v class embeddings with VGSE-SMO embeddings | SJE · CUB | ZSL Top-1 accuracy: 14.4 -> 26.1 | Helped |
| Replace w2v class embeddings with VGSE-SMO embeddings | SJE · SUN | ZSL Top-1 accuracy: 26.3 -> 35.8 | Helped |
| Add class-discrimination loss (ablation) | VGSE patch-clustering model · CUB | ZSL Top-1 accuracy: 16.7 -> 23.7 | Helped |
| Add semantic-relatedness loss (ablation) | VGSE patch-clustering model · SUN | ZSL Top-1 accuracy: 30.5 -> 35.8 | Helped |
| Replace learned patch clustering with pretrained ResNet patch features (ablation) | VGSE-SMO · CUB | ZSL Top-1 accuracy: 15.4 -> 26.1 | Hurt |
| Replace learned patch clustering with k-means clustering (ablation) | VGSE-SMO · CUB | ZSL Top-1 accuracy: 15.0 -> 26.1 | Hurt |
| Replace SMO with weighted averaging of neighbor seen-class embeddings (ablation) | VGSE · AWA2 | ZSL Top-1 accuracy: 57.7 -> 62.4 | Hurt |
| Replace SMO with weighted averaging of neighbor seen-class embeddings (ablation) | VGSE · CUB | ZSL Top-1 accuracy: 25.8 vs 26.1; < 0.5% | No clear effect |
| Use 128 patches per image instead of 9 (ablation) | VGSE-SMO · AWA2 | ZSL accuracy: 62.4% -> 58.7% | Hurt |
| Use watershed regions instead of regular 3×3 grid patches (ablation) | VGSE-SMO · AWA2 | ZSL accuracy: +8.2% | Helped |
| Increase the visual-cluster count up to 200 (ablation) | VGSE-SMO · AWA2 | ZSL accuracy: 58.4% to 62.5% | Helped |
Closest papers by meaning
- Adaptive and Generative Zero-Shot Learning (ICLR 2021 · similarity 0.93)
- ZeroDiff: Solidified Visual-semantic Correlation in Zero-Shot Learning (ICLR 2025 · similarity 0.91)
- Progressive Semantic-Guided Vision Transformer for Zero-Shot Learning (CVPR 2024 · similarity 0.90)
- ZeroMamba: Exploring Visual State Space Model for Zero-Shot Learning (AAAI 2025 · similarity 0.90)
- Class Semantic Attribute Perception Guided Zero-Shot Learning (AAAI 2025 · similarity 0.90)
- Improving Generalized Zero-Shot Learning by Exploring the Diverse Semantics from External Class Names (CVPR 2024 · similarity 0.89)
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.