Video Emotion Recognition 벤치마크
Video Emotion Recognition on CREMA-D
Accuracy
- 2023-05-12 — VAVL: Accuracy 82.6
- 2024-12-25 — MTCAE-DFER: Accuracy 85.03
- 2025-05-05 — VAEmo: Accuracy 85.68
| Rank | Model | Accuracy | WAR | Paper | Code | Year |
|---|---|---|---|---|---|---|
| 1 | VAEmo | 85.68% | – | VAEmo: Efficient Representation Learning for Visual-Audio Emotion with Knowledge Injection | MSA-LMC/VAEmo | 2025 |
| 2 | MTCAE-DFER | 85.03% | – | MTCAE-DFER: Multi-Task Cascaded Autoencoder for Dynamic Facial Expression Recognition | Peihao-Xiang/MTCAE-DFER | 2024 |
| 3 | VAVL | 82.60% | – | Versatile audio-visual learning for emotion recognition | 2023 | |
| 4 | MultiMAE-DER | – | 79.36% | MultiMAE-DER: Multimodal Masked Autoencoder for Dynamic Emotion Recognition | Peihao-Xiang/MultiMAE-DFER | 2024 |