Dataset · mmm8383
795 hours of first-person video from Project Aria glasses
mmm8383/NextMe-800
114 days of one person's daily life with video frames, gaze tracking, audio, and activity annotations.
What's inside
Contains 527 recordings (795 hours) of one person's daily life captured on Project Aria glasses in 2026 in English and Chinese. Includes 2880x2880 RGB frames at ~1 fps, gaze direction data, audio with transcripts, detailed text descriptions, and a five-level activity hierarchy. Also features the NextAct benchmark of 1,500 data points for next-action prediction.
What you can do with it
- Training multimodal video-language models (VLMs) to understand first-person actions
- Developing assistants that predict a user's next steps from video and gaze
- Evaluating models on the NextAct behavior prediction benchmark
- Tasks
- Video text → text, Text generation
- Languages
- English, Chinese
- Size
- 1M–10M rows
- Downloads this month
- 2.3K
- Likes
- 4
- Access
- on request on Hugging Face
How to load it
from datasets import load_dataset
ds = load_dataset("mmm8383/NextMe-800")