LITCHI EGO

A head-mounted first-person data collection device that turns what you see and do into training data. It records RGB × 4, depth, IMU and audio in sync, and with the T1 Pro data collection glove captures hand motion on the same timeline.

Record every part of the task from the human point of view.

LITCHI EGO is a head-mounted data collection device developed by LITCHI. A hybrid of four wide-angle RGB cameras and one structured-light stereo depth camera records both hands, the body and the workspace from a first-person view.

Hybrid of 4 RGB cameras + 1 depth camera
Built-in 9-axis IMU (500 Hz) and four microphones
Under 400 g, continuous collection with magnetic swappable batteries

Recommended: the EGO + T1 Pro set

EGO records what the operator saw and where they worked; T1 Pro records how the hands moved and where and how hard they touched. The two devices share one timeline with sync error under 1 ms, producing data ready for robot imitation learning.

Recorded by EGO	First-person RGB video (×4), depth, head motion (9-axis IMU), audio
Recorded by T1 Pro	Finger joint angles, 3D hand position and orientation and synchronized two-hand motion from 11 IMUs + tactile data from 368 points across the palm
Combined	Records “look → reach → grasp → apply force” on one timeline, with no alignment needed in post-processing

Training data for imitation learning and VLABuild data that teaches humanoids and dual-arm robots human seeing and hand motion together.

Recording contact-rich tasksRecords where and how much force is applied when grasping, pressing or assembling. For assembly, food processing, lab work and more.

Stable near robotsT1 Pro uses no magnetic sensors, so it stays stable near motors and high-current wiring, which suits data collection alongside teleoperation.

* If you would like us to handle everything from setting up collection to processing and annotation, our data collection service is also available. Contact HumansX to discuss configurations.

What a first-person view can record

Wide, deep and precise: designed for first-person data

Field of view, depth, motion and sync: everything training data needs, in one head-mounted device.

Wide field of viewEach RGB camera covers 165° horizontal / 130° vertical. Four cameras record from the hands to the surrounding workspace with few blind spots.

最大 1920 × 1200 @ 30 fps
Global shutter

Depth sensingA structured-light stereo depth camera precisely captures object position, distance and approach in close-range work within 1.5 m.

空間精度 < 0.5%（1 m）
1280 × 800 @ 30 fps、H 90° / V 65°

Motion recordingGlobal shutters suppress distortion and blur even in fast motion. A 9-axis IMU records head motion at the same time.

9軸 IMU 500 Hz
Image–IMU sync < 1 ms

Multi-device syncRecords first-person video, depth, inertial data and hand motion on one timeline, reducing alignment work in multimodal post-processing.

EGO × T1 Pro 同期 < 1 ms
External trigger input supported

Stable through long and repeated collection sessions

Light, stable fitUnder 400 g with a self-adapting structure that fits any wearer and reduces fatigue over long sessions. Stable extrinsics keep collection conditions consistent.

Self-calibrationSelf-calibration for long-term use recalibrates camera–sensor offsets and suppresses parameter drift.

Non-stop continuous collectionMagnetic swappable batteries keep collection running during battery changes, for long sessions without interruption.

One-touch collection managementStart and stop collection with one button. Sessions are auto-named and saved per task, and data can be retrieved over USB or wirelessly.

From the human view to model capability

Building real-world datasets: large-scale recording of real work across regions, industries and tasks
Multimodal and physical world-model training: synchronized vision, depth, inertial and audio data
Research on human work behavior: reproduce observation, movement and manipulation in real environments
Robot manipulation learning: combined with T1 Pro as a data foundation for imitation and task learning

Product page ↗Request materials