Hardware platform for collecting embodied AI and robot learning training data without a robot body. Modules include egocentric cameras, RGB-D, a wrist camera and UMI finger/gripper units.
Handheld gripper kit that records manipulation demonstrations with GoPro HERO13 Black cameras and outputs Zarr or MCAP datasets for diffusion-policy training.
GenAI data augmentation service that converts recorded test-drive video into variants across weather and lighting conditions, adding or removing objects while keeping existing annotations.
Perception training datasets (traffic lights, traffic signs, pedestrians, vehicles, highway) for camera, LiDAR, and radar, with 2D/3D bounding boxes and semantic and instance segmentation.
Browser-based agentic simulation co-pilot in Falcon that turns natural-language prompts into scenarios and synthetic datasets with ground truth.
Library of validated 3D assets and scenes for robot simulation training, with commercial licenses, accessed via simready.com.
Egocentric human data solution capturing RGB-D, upper-body and hand pose, and tactile sensor data across 7 countries. Over 300,000 hours of data delivered.
Large-scale dataset of first-person human interaction data from real-world activity across tasks, spaces and environments, for robot learning and human-to-robot transfer.
iOS app for capturing environment-specific and task-specific data at scale.
Robot data offered as open-source trajectories, standard datasets for tasks and scenes, and custom data collection pipelines.
Video, navigation, and urban mapping data collected on daily robot routes, covering street conditions, accessibility checks, and infrastructure alerts, for business partners and cities.
Fully managed cloud platform generating annotated synthetic image data with physics-based rendering and sensor simulation to train and validate Vision AI models.
Robotic motion dataset generation service using multi-joint robots, with optional data cleansing and curation. Launched January 2026.