Skip to content

EgoLife

A project dataset for long-term multimodal daily-life AI research

Visit official site

What EgoLife is

EgoLife is described as an AI assistant project aimed at long-term, multimodal and multi-view daily life, with six volunteers recorded over one shared week into about 50 hours of video. The project also proposes the EgoRAG framework for long-term context understanding tasks.

What you can do with it

  • Analyse long-term events across hours and days
  • Work with synchronised video, gaze and IMU data
  • Use 15 GoPro cameras and first-person glasses for multi-view recording
  • Use 3D scans, transcripts and dense captions in model training

Example uses

  • Researchers train AI models to follow events in long videos
  • Developers build video analysis tools for smart homes or health monitoring
  • Educators study human behaviour with social interaction data

Getting started

  • Visit the project site to read about the background and datasets
  • Download the video, 3D scan and annotation data
  • Use the EgoLifeQA benchmark to evaluate long-term context tasks
  • Apply the EgoRAG framework to develop or refine multimodal models

Pros & cons

✓ What we like

  • Long-term event analysis across hours and days is supported
  • Transcripts, dense captions and 3D scans are provided

! What to watch out for

  • The described output is a research dataset and benchmark
  • The recording covers one shared week of six volunteers

FAQ

What data is included?

The material lists video, gaze and IMU data, 3D scans, transcripts and dense captions.

What is EgoRAG?

The material describes a framework for long-term context understanding tasks.

Last reviewed: 2026-11-15

More AI office assistant tools

View all →

How we review