Loading...

Synthetic data for

Robotics

Robots are transforming how we live and work from manufacturing and industrial operations, to residential cleaning and cooking, to public safety and security inspections.

Applications
Inventory monitoring
Human-robot interactions
Safety hazard detection
Deliveries
And much more

Empowering these use cases

Never be blocked by lack of data. Build computer vision models for these use cases (and more!).

Inventory monitoring

Efficiently monitor inventory on shelves by having a robot drive through aisles or on shelves.

Human-robot interactions

Build robots that seamlessly interact with humans by understanding behaviors like gestures, distracted body language, or following a human when walking.

Safety hazard detection

Automatically detect safety hazards like spills, falls, and medical issues. Patrol buildings for security events like weapons, breached perimeters, and tailgating

Data that meets your exact specs

Generate robotics datasets that meet your specs by selecting distributions of objects, scenes, and human-robot interactions. Use the API to define the specs that matter and randomize the rest.

Pixel-perfect labels included

Every video comes with pixel-perfect labels including depth maps, camera matrices and lens parameters, scene metadata (incl. robot speed), activity labels, cuboids, segmentation masks, and much more.

Simulate any robot

Generate synchronized videos across multiple cameras in any configuration. Simply pick the number of cameras, their lens types, and their positions.

Features

Build better models faster with feature-rich synthetic data from the Robotics API.

Any lens type

Choose your camera lens type from fisheye to perspective. Set focal length, field of view, and sensor size parameters. The API enables users to generate videos of the same scene from multiple different lens types.

Set robot speed and camera height

Select the speed of the robot and the smoothness of its trajectory in the facility. The API enables users to generate videos of the same scene from multiple different robot speeds and camera heights.

Synchronized cameras

Get synchronized videos from an arbitrary number of cameras. Cameras can be at different angles and heights from each other.

Import scenes

Choose scenes from the Infinity library or import your own into 3D simulation. Specific data for a facility is especially useful in cases where models must be kept small (such as edge deployment) or failure cases addressed. Learn more

Build specific object classifiers

Build specific object classification models by importing specific objects into 3D. Procedurally place those objects into the scene, such as specific product inventory items, ladders or tools, PIT vehicles, and more. Get ground-truth bounding boxes and segmentation masks for any object.

Resources



Book an intro call to get started or try out a synthetic dataset today

Schedule a 30-min time-slot with us directly, or send us a message if there's something else we can help with.

During the call, we'll show you a live demo, discuss your needs, and identify how we can help.

For a look ahead after the call, take a look at our API and data flywheel technologies.