Ego-Exo Dataset

Combines first-person wearable camera views with external camera perspectives for comprehensive spatial understanding. Multi-camera synchronization captures both the user’s viewpoint and surrounding context.

Key Features

Ego + Exo Views

First-person and external views in a single synchronized dataset

Camera Sync

Timestamp QR-code for millisecond-level multi-device alignment

Interaction Focus

Both object manipulation and human-human interaction scenarios

Real-World Tasks

Everyday activities in authentic environments for practical AI training

Data Collection Setup

Multi-Camera Synchronized Collection

One wearable camera and two external cameras, synchronized via timestamp QR-code technology for precise alignment across all three viewpoints.

Camera Setup	1 wearable + 2 external cameras
Views	Ego view + External views
Sync	Timestamp QR-code
Data Type	Multi-view video dataset
Environment	Real-world scenarios

Multi-View Data Structure

Three-Camera Perspective System

Ego View

From the user’s perspective
Hand movement details
Object interaction process
First-person action information

Exo View

From external cameras
Full-body movement
Spatial relationships
Environmental context

Dataset Examples

Example 1: Clothing Folding

A human performs clothing folding in a real environment, captured from three synchronized viewpoints.

First-person wearableFront external cameraSide external camera

Captures: Hand-object interaction · Folding process · Manipulation behavior

Ego View

Exo View1

Exo View2

Example 2: Human Care Activity

Multi-view recordings of human care activities covering 9 body regions.

BackShoulderChestWristArmLegNeckHeadFoot

Captures: Human motion patterns · Body interaction process · Multi-view spatial info

Ego View

Exo View1

Exo View2

Example 3: Baking — Cake Making

A human makes a cake in a kitchen environment, captured from three synchronized viewpoints.

First-person wearableFront external cameraSide external camera

Captures: Ingredient handling · Batter mixing & pouring · Icing/decorating · Tool-object manipulation process · Multi-view spatial info

Ego View

Exo View1

Exo View2

Example 4: Haircut

A stylist performs a haircut on a client in a salon environment, captured from three synchronized viewpoints.

First-person wearableFront external cameraSide external camera

Captures: Tool-hair interaction · Cutting/styling technique · Two-person spatial coordination

Ego View

Exo View1

Exo View2

Example 5: Beauty / Skincare

A beautician performs a facial skincare routine on a client, captured from three synchronized viewpoints.

First-person wearableFront external cameraSide external camera

Captures: Fine hand-face interaction · Product application technique · Multi-view facial detail

Ego View

Exo View1

Exo View2

Dataset Specifications

Data Type	Multi-view Video Dataset
Camera Config	1 Ego + 2 Exo Cameras
Sync	Timestamp QR-code
Perspective	First-person + Third-person
Environment	Real-world scenarios
Tasks	Manipulation and interaction
Annotation	Available upon request

Applications

Supporting Advanced AI Research

Humanoid Robot Learning Robot Manipulation VLA Models Human-Robot Interaction Multimodal AI Embodied Intelligence

Interested in using this dataset for your research or product development? Get in touch with our team.