Research map / Label-efficient and robust learning / Recognition and long-tail
Context Based Emotion Recognition using EMOTIC Dataset
Ronak Kosti, José Manuel Álvarez, Adrià Recasens, Àgata Lapedriza
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 whole-image (scene context) branch to body branch (ablation) | CNN model with Smooth L1 · EMOTIC discrete categories | mean AP (%): 22.36 -> 27.38 | Improved |
| add whole-image (scene context) branch to body branch (ablation) | CNN model with Euclidean loss · EMOTIC discrete categories | mean AP (%): 23.86 -> 24.88 | Improved |
| add context branch to body branch (ablation) | CNN model · EMOTIC continuous dimensions | mean AAE: 0.0569 -> 0.0573 | No clear effect |
| use Smooth L1 instead of Euclidean continuous loss with body+image (ablation) | B+I · EMOTIC discrete categories | mean AP (%): 24.88 -> 27.38 | Improved |
| use Smooth L1 instead of Euclidean continuous loss with body only (ablation) | B · EMOTIC discrete categories | mean AP (%): 23.86 -> 22.36 | Hurt |
| add context to body for rarely-annotated negative emotions (ablation) | B+I (SL1) · EMOTIC Sadness | AP (%): 08.13 -> 19.66 | Improved |
| add context to body (ablation) | B+I (SL1) · EMOTIC Suffering | AP (%): 04.90 -> 18.84 | Improved |
| add context to body (ablation) | B+I (SL1) · EMOTIC Happiness | AP (%): 58.71 -> 58.26 | No clear effect |
| add Places-finetuned image features If to body features in logistic regression (ablation) | Bf · EMOTIC discrete categories | mean AP (%): 23.00 -> 27.70 | Improved |
| use sentiment-score features Is instead of fine-tuned Places features If (ablation) | Bf+context · EMOTIC discrete categories | mean AP (%): 27.70 -> 29.45 | Improved |
| use sentiment-score features Is instead of If (ablation) | Bf+context · EMOTIC continuous dimensions | mean AAE: 0.0643 -> 0.0713 | Hurt |
Closest papers by meaning
- Emotion Recognition in Context (CVPR 2017 · similarity 0.94)
- EmoNet-Face: An Expert-Annotated Benchmark for Synthetic Emotion Recognition (NeurIPS 2025 · similarity 0.88)
- AffectGPT: A New Dataset, Model, and Benchmark for Emotion Understanding with Multimodal Large Language Models (ICML 2025 · similarity 0.87)
- FindingEmo: An Image Dataset for Emotion Recognition in the Wild (NeurIPS 2024 · similarity 0.87)
- OUS: Bridging Scene Context and Facial Features to Overcome the Rigid Cognitive Problem (AAAI 2025 · similarity 0.87)
- OV-MER: Towards Open-Vocabulary Multimodal Emotion Recognition (ICML 2025 · similarity 0.87)
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.
