Cleaning windows12:00
Cleaning windows with pole
105 actions40 skillsMCAP 11.6 GB
Real-world egocentric data recorded on the job: egocentric and wrist video, depth, IMU, hand pose and frame-level action labels in one MCAP per clip. Free to use with attribution.
120.4 h5,392 clips12 categories7.3 TBCC-BY-4.0
Categories12
Browse all clips
21 clips
Cooking food0:18
Frying Onions
4 actions7 skillsMCAP 285 MB
Mopping floor15:43
Mopping office floor
72 actions28 skillsMCAP 16.1 GB
Organizing items11:07
Pulling room curtains
93 actions40 skillsMCAP 7.9 GB
Preparing cleaning supplies1:12
Preparing mop bucket
13 actions17 skillsMCAP 1.2 GB
Sweeping floor0:20
Sweeping debris into dustpan
2 actions6 skillsMCAP 271 MB
Wiping surfaces1:35
Cleaning table surface
10 actions9 skillsMCAP 1.2 GB
Cleaning windows3:58
Cleaning building windows
33 actions15 skillsMCAP 4.8 GB
Mopping floor7:22
Mopping hallway floor
60 actions40 skillsMCAP 6.4 GB
Preparing cleaning supplies0:23
Wringing out mop head
3 actions9 skillsMCAP 318 MB
Cleaning windows0:40
Cleaning building windows
6 actions9 skillsMCAP 814 MB
Mopping floor10:30
Mopping office floor
52 actions20 skillsMCAP 9.7 GB
Mopping floor6:50
Mopping office floor
42 actions24 skillsMCAP 6.0 GB
Mopping floor4:59
Mopping stairs and landing
32 actions24 skillsMCAP 4.4 GB
Mopping floor5:10
Mopping hallway floor
32 actions30 skillsMCAP 4.4 GB
Mopping floor5:51
Mopping floor
31 actions30 skillsMCAP 6.1 GB
Mopping floor1:17
Mopping floor
11 actions9 skillsMCAP 1.1 GB
Mopping floor0:56
Mopping tiled floor
11 actions4 skillsMCAP 806 MB
Mopping floor0:33
Mopping the floor
6 actions6 skillsMCAP 478 MB
Mopping floor1:08
Mopping entrance floor
6 actions5 skillsMCAP 1.1 GB
Mopping floor0:20
Mopping lecture hall floor
2 actions6 skillsMCAP 379 MB
In every MCAP
23 channels · one clock · ROS 2| Topic | Schema | Rate |
|---|---|---|
| /action | tad_msgs/msg/ActionPhase | 30 Hz |
| /ego/zed_head/depth/camera_info | sensor_msgs/msg/CameraInfo | 30 Hz |
| /ego/zed_head/depth/image/compressed | sensor_msgs/msg/CompressedImage | 30 Hz |
| /ego/zed_head/imu/data | sensor_msgs/msg/Imu | 199.8 Hz |
| /ego/zed_head/left/camera_info | sensor_msgs/msg/CameraInfo | 30 Hz |
| /ego/zed_head/left/image/compressed | foxglove.CompressedVideo | 30 Hz |
| /ego/zed_head/odom | nav_msgs/msg/Odometry | 30 Hz |
| /ego/zed_head/right/camera_info | sensor_msgs/msg/CameraInfo | 30 Hz |
| /ego/zed_head/right/image/compressed | foxglove.CompressedVideo | 30 Hz |
| /hands/left/detected | std_msgs/msg/Bool | 30 Hz |
| /hands/left/keypoints | visualization_msgs/msg/MarkerArray | 30 Hz |
| /hands/left/odom | nav_msgs/msg/Odometry | 30 Hz |
| /hands/right/detected | std_msgs/msg/Bool | 30 Hz |
| /hands/right/keypoints | visualization_msgs/msg/MarkerArray | 30 Hz |
| /hands/right/odom | nav_msgs/msg/Odometry | 30 Hz |
| /tf | tf2_msgs/msg/TFMessage | 30 Hz |
| /tf_static | tf2_msgs/msg/TFMessage | 2.9 Hz |
| /wrist_left/image/camera_info | sensor_msgs/msg/CameraInfo | 30 Hz |
| /wrist_left/image/compressed | foxglove.CompressedVideo | 30 Hz |
| /wrist_left/imu/data | sensor_msgs/msg/Imu | 199.4 Hz |
| /wrist_right/image/camera_info | sensor_msgs/msg/CameraInfo | 30 Hz |
| /wrist_right/image/compressed | foxglove.CompressedVideo | 30 Hz |
| /wrist_right/imu/data | sensor_msgs/msg/Imu | 200 Hz |
Get the data
EgoViz-120 is 7.3 TB, served from a public Hugging Face Storage Bucket. Start with the index: one row per clip with category, title, skills, objects and paths, before you download any video.
Install and index
pip install -U huggingface_hub
hf auth login
B=hf://buckets/humaidtech/EgoViz-120
hf buckets cp $B/index.parquet index.parquetOne category
hf sync $B/data/cleaning_kitchen ./cleaning_kitchenPreviews and metadata only
hf sync $B/data ./egoviz --include "*.mp4" --include "*.json"Cite
@misc{egoviz120_2026,
title = {EgoViz-120: 120 Hours of Synchronized Real-World Egocentric Data},
author = {Humaid Tech Inc.},
year = {2026},
publisher = {Hugging Face},
howpublished = {\url{https://huggingface.co/datasets/humaidtech/EgoViz-120}}
}