Seat cleaning · RGB and depth
See the same seat-cleaning task through two RGB cameras, recorded depth, and a point cloud calculated from that depth.
DATA FOR ROBOTICS & PHYSICAL AI
Human actions,
robot learning data.
Catabricks collects and reviews real human task and sensor data for robotics and Physical AI teams. Explore datasets or discuss a collection tailored to your task.
Captured footage · excerpt
Seat cleaning · RGB and depth
Video 1 / 148 sec
See the same seat-cleaning task through two RGB cameras, recorded depth, and a point cloud calculated from that depth.
Top: two RGB camera views · Bottom: depth and point cloud
Capture and processing details
The two views are aligned by recorded host time. The point cloud is calculated from camera 1’s depth measurements.
Explore more tasks
All 14 videosCargo cleaning · two cameras
Floor cleaning · two cameras
Cleaning wheel spokes
Cleaning beside the seat
Recorded in the field. Prepared for learning.
Explore captured tasks01 / IN THE FIELD
Recorded examples of vehicle care
Every video brings RGB, depth and point clouds together. Open a larger view to explore the task and the shape of its surroundings.
Cleaning a floor mat
A nozzle follows the shape and edges of the floor mat.
Cleaning wheel spokes
Hands work a cloth into the spaces between wheel spokes.
Wiping the driver area
A cloth follows the shapes around the steering wheel and center console.
Cleaning beside the seat
One hand braces against the seat while the other guides a nozzle across the floor and into gaps.
Wiping a vehicle surface
A short excerpt of a hand moving a cloth along a vehicle surface.
Wiping a seat
A cloth changes position to follow the contours of the seat.
Cleaning the cargo area
Cleaning the cargo area, then lifting its floor panel.
Wiping the cargo opening
A cloth wipes along the inner panel and edges of the cargo opening.
Wiping a side mirror
The cloth changes position and direction around a side mirror.
Unfolding a cloth with both hands
Both hands spread and grip the cloth before wiping the bodywork.
Cleaning around door controls
A nozzle cleans the gaps around the door handle and switches.
Floor cleaning · two cameras
Two viewpoints of the same floor-cleaning task. Follow the nozzle across the mat and the worker’s hand movements.
Cargo cleaning · two cameras
Two viewpoints show cargo-area cleaning and the floor panel being opened. See the tool movements alongside the surrounding space.
02 / BEYOND THE FRAME
Data for your training task
Video is the starting point. We define the sensors and data you need around your learning goals.
01RGB
The context of an action
Video records how hands interact with objects, the sequence of a task and the surrounding environment.
02DEPTH
The structure of a space
Depth is recorded alongside RGB. Measured distances become a point cloud for inspecting spatial structure.
03IMU
The dynamics of motion
Collections with an IMU record sensor motion, including acceleration and rotation.
Sensors and delivered data vary by collection setup and dataset.
Explore data services03 / FROM TASK TO DATASET
From requirements to data delivery
Know what is captured and how it can be used. Clear from collection to delivery.
- 01 / DEFINE
Define requirements
Define the tasks, collection environments, sensors and dataset size, then review the scope we can deliver.
- 02 / CAPTURE & CURATE
Collect and review
Record task and sensor information, then review data quality, formats and usage rights.
- 03 / DELIVER
Deliver the dataset
Deliver datasets with defined contents and licenses through the web and CLI. Verify file versions and integrity.
Explore on the web. Download to your workspace with the CLI.
CLI guideYOUR NEXT DATASET STARTS HERERequest custom data
What are you teaching your model?
Tell us your target tasks, sensors, dataset size and timeline.