Skip to content

Dataset · HumanBehaviorAtlas

Multimodal dataset for behavior and emotion analysis

HumanBehaviorAtlas/human_behavior_atlas

Combined corpus of 16 datasets with video, audio, and text for recognizing emotions, humor, and human state.

What's inside

It contains between 100 thousand and 1 million examples with video, audio, texts, and pre-extracted features in English and Chinese. The dataset combines 16 original sets into a unified schema for analyzing emotions, sentiment, humor, mental state, and video-based question answering. This data was used to train the OmniSapiens base model.

What you can do with it

  • Training multimodal models for emotion and social signal recognition
  • Testing video question-answering and speech sentiment analysis systems
Tasks
Video classification, Audio classification, Text classification, Question answering, Visual question answering, Multimodal
Languages
English, Chinese
Size
100K–1M rows
Downloads this month
20K
Likes
4

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("HumanBehaviorAtlas/human_behavior_atlas")