Research map / Label-efficient and robust learning / Recognition and long-tail
A Compact Embedding for Facial Expression Similarity
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 |
|---|---|---|---|
| learn an embedding from human-rated expression-comparison triplets rather than emotion-label supervision | 16-dimensional facial expression embedding · FEC test set | triplet prediction accuracy: 53.3 -> 81.8 | Helped |
| add weak rater-agreement triplets to training (ablation) | FECNet-16d · FEC test set | triplet prediction accuracy: 81.8 -> 80.5 | Hurt |
| use 16 embedding dimensions rather than lower-dimensional alternatives (ablation) | FECNet · FEC test set | triplet prediction accuracy | Helped |
| use features from the Inception 4e block rather than 4d, 5a, or 5b (ablation) | FECNet · FEC test set | triplet prediction accuracy | Helped |
| use human similarity comparisons, including within-category comparisons, instead of deriving similarity from emotion categories | facial expression embedding · FEC test set | triplet prediction accuracy: 53.3 -> 81.8 | Helped |
| train the embedding with human preference triplets rather than classification objectives | emotion classification using FECNet-16d + K-NN · AffectNet validation set | average AUC-ROC: 88.6 -> 82.9 | Hurt |
Closest papers by meaning
- When Facial Expression Recognition Meets Few-Shot Learning: A Joint and Alternate Learning Framework (AAAI 2022 · similarity 0.91)
- EmotioNet: An Accurate, Real-Time Algorithm for the Automatic Annotation of a Million Facial Expressions in the Wild (CVPR 2016 · similarity 0.90)
- Multi-Dimensional, Nuanced and Subjective - Measuring the Perception of Facial Expressions (CVPR 2022 · similarity 0.89)
- From Pixels to Semantics: Unified Facial Action Representation Learning for Micro-Expression Analysis (ICLR 2026 · similarity 0.88)
- EmoNet-Face: An Expert-Annotated Benchmark for Synthetic Emotion Recognition (NeurIPS 2025 · similarity 0.86)
- A Unified and Interpretable Emotion Representation and Expression Generation (CVPR 2024 · similarity 0.86)
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.