Eidon AI

Eidon built a wearable rig for capturing human manipulation: a head-mounted camera plus a seven-point IMU harness. We ran it across ordinary household work like laundry, cleaning, dishes and cooking.

The company has wound down. Everything we collected is published here under CC-BY-4.0, so it outlives us.

How the release is organised

The data sits in three places. They are parts of one corpus rather than versions of it, and most people will want the first two together.

What's inside Size Load with
tracker-pov The videos. 13,451 egocentric MP4s and one metadata row each 9.05 TB load_dataset
tracker-pov-imu The sensors. 779M rows of 24 Hz orientation and motion for the same recordings 9.5 GB load_dataset
egocentric-pov Extra egocentric video with no sensor data 1.55 TB hf buckets sync

The third one is a Storage Bucket rather than a dataset, so it will not show up in the Datasets list above and load_dataset cannot read it. Reach it with hf buckets sync or the S3-compatible API.

The first two are one dataset in two pieces

tracker-pov and tracker-pov-imu cover the same 13,451 recordings, joined on recording_id. Take the video, the sensors, or both.

from datasets import load_dataset

video = load_dataset("eidon-ai/tracker-pov", split="train")                  # 9.05 TB
imu   = load_dataset("eidon-ai/tracker-pov-imu", split="train", streaming=True)  # 9.5 GB

Filter the metadata first. It is small, and it carries the QC scores, task labels and sensor coverage flags you will want to select on:

import pandas as pd
meta = pd.read_parquet("hf://datasets/eidon-ai/tracker-pov/recordings/metadata.parquet")
good  = meta[(meta.qc_status == "valid") & (meta.has_chest)]

What's in it

1,274 hours of paired egocentric video and seven-point arm tracking, from 27 contributors in their own homes. Every recording passed an automated QC pass for hand presence, lighting, blur and camera stability, and ships with its scores.

The bucket holds another 306 hours of egocentric video across 1,370 recordings, from 37 contributors, 14 of whom also appear in the tracker set and keep the same contributor_id. Same rig and same QC pipeline, without the IMU harness. Being a bucket, it is downloaded rather than loaded: hf buckets sync hf://buckets/eidon-ai/egocentric-pov ./egocentric-pov.

Read the limitations section before training on it. The short version: the task distribution leans heavily toward folding laundry, contributors are few and unevenly weighted, and you should split by contributor_id rather than randomly.

Licence and contact

CC-BY-4.0. Commercial use allowed, attribution required. Published by Solidic Labs Inc (Eidon AI).

For removal requests or questions: padilla.samuelk@gmail.com