Emotion Recognition in Context on EMOTIC (test)
35.48mAPEmotiCon
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| EmotiConMain idea=Multimodal context with personal, semantic, and socio-dynamic cues2026.06 | 35.48 | — | |
| CLIP-based two-stream model (Ours)Main idea=Image-only ResNet-18 body stream + CLIP scene stream, backbone=ResNet-182026.06 | 34.52 | — | |
| Kosti et al.Main idea=Person crop and scene CNN fusion for emotion recognition in context2026.06 | 27 | — |