Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Zero-Shot Emotion Recognition via Affective Structural Embedding
Chi Zhan, Dongyu She, Sicheng Zhao, Ming-Ming Cheng, Jufeng Yang
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 the affective adversarial constraint Ladv (ablation) | affective structural embedding · FI 6:2 | accuracy (%): 66.41 -> 67.53 | Improved |
| add Ladv to the basic embedding only (with L*as) (ablation) | Base · FI 6:2 | accuracy (%): 65.12 -> 65.46 | Improved |
| add structural alignment loss Las (ablation) | Base · FI 6:2 | accuracy (%): 65.12 -> 66.41 | Improved |
| add reconstruction loss Lre to Las (ablation) | Base + Las · FI 6:2 | accuracy (%): 65.12 -> 67.28 | Improved |
| add class-center alignment (second part of Las) (ablation) | Base + L*as · FI 6:2 | accuracy (%): 65.46 -> 66.41 | Improved |
| train all components together (ablation) | affective structural embedding · FI 6:2 | accuracy (%): 67.53 -> 68.87 | Improved |
| concatenate ANP and deep features (M) for baseline methods (ablation) | comparison ZSL methods · FI/ArtPhoto/Abstract/IAPSa/Emotion6 | accuracy (%) | No clear effect |
| use ANP features instead of deep features | baseline ZSL methods · Emotion6 | accuracy (%) | Mixed |
| use proposed method over CDL | ZSL emotion recognition · FI generalized ZSL 6:2 | accuracy (%): 35.72 -> 39.54 | Improved |
| use proposed method over CDL in cross-dataset recognition | ZSL emotion recognition · FI->Emo6 | accuracy (%): 55.15 -> 61.94 | Improved |
| change test classes to excitement and sadness | Ours · FI 6:2 | accuracy (%): 68.87 -> 74.03 | Improved |
| change test classes to all negative emotions | Ours · FI 4:4 | accuracy (%): 54.73 -> 42.92 | Hurt |
Closest papers by meaning
- Learning Visual Emotion Representations From Web Data (CVPR 2020 · similarity 0.88)
- Learning Unseen Emotions from Gestures via Semantically-Conditioned Zero-Shot Perception with Adversarial Autoencoders (AAAI 2022 · similarity 0.88)
- Retrieving and Classifying Affective Images via Deep Metric Learning (AAAI 2018 · similarity 0.87)
- MART: Masked Affective RepresenTation Learning via Masked Temporal Distribution Distillation (CVPR 2024 · similarity 0.85)
- Bridge Then Begin Anew: Generating Target-Relevant Intermediate Model for Source-Free Visual Emotion Adaptation (AAAI 2025 · similarity 0.85)
- AffectGPT: A New Dataset, Model, and Benchmark for Emotion Understanding with Multimodal Large Language Models (ICML 2025 · similarity 0.84)
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.
