Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Classifier and Exemplar Synthesis for Zero-Shot Learning
Soravit Changpinyo, Wei‐Lun Chao, Boqing Gong, Fei Sha
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 |
|---|---|---|---|
| use predicted exemplars as semantic representations (EXEM) instead of given ones | SynC struct · ImageNet (All) | ZSL per-class accuracy: 0.99 -> 1.29 | Improved |
| apply EXEM exemplar prediction to base ZSL methods | SynC, ConSE, ESZSL · small datasets | ZSL accuracy: relatively 2.1-13.3%, 11.4-32.7%, and 1.6-12.0% | Improved |
| use standardized Euclidean distance (1NNs) instead of Euclidean (1NN) | EXEM nearest neighbor · CUB (new split) | ZSL per-class accuracy: 54.2 -> 58.0 | Improved |
| use standardized Euclidean distance (1NNs) instead of Euclidean (1NN) | EXEM nearest neighbor · SUN (new split) | ZSL per-class accuracy: 63.0 -> 62.9 | No clear effect |
| use standardized Euclidean distance (1NNs) instead of Euclidean (1NN) | EXEM nearest neighbor · ImageNet (All) | ZSL per-class accuracy: 1.26 -> 1.29 | Improved |
| regularize the bases v_r instead of w_c (ablation) | SynC o-vs-o (GoogLeNet) · AwA | ZSL accuracy: 69.7% -> 71.7% | Improved |
| regularize the bases v_r instead of w_c (ablation) | SynC o-vs-o (GoogLeNet) · CUB | ZSL accuracy: 53.4% -> 56.4% | Improved |
| regularize the bases v_r instead of w_c (ablation) | SynC o-vs-o (GoogLeNet) · SUN | ZSL accuracy: 62.8% -> 67.5% | Improved |
| learn phantom class semantic representations (ablation) | SynC o-vs-o (GoogLeNet) · AwA | ZSL accuracy: 69.7% -> 71.1% | Improved |
| learn phantom class semantic representations (ablation) | SynC o-vs-o (GoogLeNet) · CUB | ZSL accuracy: 53.4% -> 54.2% | Improved |
| learn phantom class semantic representations (ablation) | SynC o-vs-o (GoogLeNet) · SUN | ZSL accuracy: 62.8% -> 63.3% | Improved |
| reduce the number of seen classes from 645 to 72 (ablation) | EXEM (1NN) · SUN (new split) | ZSL accuracy: absolute 14.4% drop | Hurt |
Closest papers by meaning
- Synthesized Classifiers for Zero-Shot Learning (CVPR 2016 · similarity 0.93)
- Predicting Visual Exemplars of Unseen Classes for Zero-Shot Learning (ICCV 2017 · similarity 0.92)
- An Empirical Study and Analysis of Generalized Zero-Shot Learning for Object Recognition in the Wild (ECCV 2016 · similarity 0.92)
- Zero-Shot Learning on Semantic Class Prototype Graph (TPAMI 2017 · similarity 0.92)
- ZeroDiff: Solidified Visual-semantic Correlation in Zero-Shot Learning (ICLR 2025 · similarity 0.90)
- Data-Free Generalized Zero-Shot Learning (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.
