atlas

Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning

Contrastive Embedding for Generalized Zero-Shot Learning

CVPR 2021 · Label-efficient and robust learning › Zero-/few-shot and meta-learning · approach: Generative feature synthesis for zero-shot · 286 citations (OpenAlex)

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 changedWhereMeasuredEffect
Replace semantic-space classification with a new embedding space (ablation)Gen+SE hybrid · AWA1 GZSLH: 55.9 -> 66.9Helped
Use contrastive embedding instead of semantic embedding in the new embedding space (ablation)Feature-generation hybrid · CUB GZSLH: 63.1 -> 65.3Helped
Add class-level contrastive supervision (ablation)CE-GZSL · AWA1 GZSLH: 67.8 -> 69.1Helped
Add instance-level contrastive supervision (ablation)CE-GZSL · CUB GZSLH: 63.0 -> 65.3Helped
Increase embedding dimension from 256 to 2,048 (ablation)CE-GZSL · AWA1 GZSLH: 53.3 -> 69.0Helped
Increase synthesized examples per unseen class (ablation)CE-GZSL · Five GZSL datasetsHHelped
Vary contrastive temperatures (ablation)CE-GZSL · Five GZSL datasetsHNo clear effect
Use a large random mini-batch instead of hand-crafted positive/negative sampling (ablation)CE-GZSL · AWA1 GZSLH: 68.2 -> 69.1Helped

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy