New SpEmoC dataset targets balanced multimodal emotion recognition

Researchers have introduced SpEmoC, a new benchmark dataset designed for multimodal emotion recognition in spoken conversations. The dataset comprises over 300,000 clips from movies and TV series, with a curated subset of 30,000 high-quality clips featuring synchronized visual, audio, and textual modalities. SpEmoC aims to address limitations in existing datasets by providing a balanced distribution of seven emotions, including minority classes like fear and disgust, and employing strict data splitting to ensure reliable evaluation of model generalization. AI

IMPACT This dataset could improve the robustness and generalizability of AI models for understanding human emotions in various applications.

RANK_REASON The item describes a new academic dataset and benchmark for multimodal emotion recognition, published on arXiv. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• disgust
• fear
• Gotit.pub
• Hugging Face
• Santosh Vipparthi Kumar
• ScienceCast
• SpEmoC

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →