What’s Out There
← Back to swiping
Images & videoBy applicationAcademic

Ego4D

Ego4D consortium (Meta AI and university partners)

3,670 hours of daily life filmed from head-mounted cameras by 923 people in 9 countries.

First-person video of ordinary activities such as cooking, shopping, and socializing, recorded by participants wearing cameras, with some audio, gaze, 3D scans, and narrations. It underpins benchmarks for episodic memory ('where did I leave my keys?'), hand-object interaction, and social understanding, and is the leading data set for AR-glasses research.

Access

Cost
By applicationFree, but users must sign a license agreement and be approved; access has moved to Lab22.
License
Open, non-commercial onlyEgo4D license agreement (research use)
Sign-up
Application required
How you get it
Bulk downloadMP4, JSON

What's in it

One row is
One video clip from a participant's head-mounted camera, with narrations and benchmark annotations
Covers
74 locations in 9 countries
Years
Released 2021-2022
Updated
Irregular
Size
3,670 hours of video from 923 participants
videoparticipant IDlocationscenarionarrationsbenchmark annotations

Quality

5 of 5

Unprecedented scale for first-person video, with consent and privacy policies at each site and versioned benchmarks.

  • Requires an approved license agreement
  • Recording practices and redaction vary by partner site
  • Multi-terabyte download

Connects to

  • Kinetics-700: Third-person action clips to contrast with first-person footage
  • AudioSet: Sound-event labels useful for Ego4D's audio-visual tasks

Questions it could answer

  • How much of a typical day is spent on each activity scenario across countries?
  • How quickly can a model find the last time an object was seen in hours of video?

Similar data sets

  • OpenRoboCare

    315 sessions of occupational therapists doing caregiving tasks, with video, touch, and gaze.

    Images & videoEmPRISE Lab at Cornell UniversityFree

  • AgiBot World 2026

    Real-world robot demonstrations from homes and commercial spaces, collected on the AgiBot G2 platform.

    Images & videoAgiBotFree

  • Kinetics-700

    650,000 ten-second YouTube clips of people doing 700 things, from yoga to changing a tire.

    Images & videoGoogle DeepMind (hosted by the Common Visual Data Foundation)Free

  • Waymo Open Dataset

    Lidar, camera, and motion logs from real robotaxis, released for anyone doing research.

    Images & videoWaymoFree

  • FineART

    40,543 robot episodes and 533,913 human-written subtask labels covering 1,718 hours.

    Images & videoScale AI and Hugging Face LeRobotFree

  • PhysicalAI Autonomous Vehicles

    About 1,700 hours of multi-sensor driving data from diverse geographies for end-to-end driving research.

    TransportationNVIDIAFree