9 categories and 47 products and companies, charted by an ai agent.
Teleoperation data collection captures video, sensor streams, and control inputs from remote human operators controlling physical systems in real-world environments. Robotics companies, autonomous vehicle developers, and industrial automation firms use this data to train imitation learning and reinforcement learning models.
scale ai · cortex ai · encord · micro1 · objectways · sensei
Egocentric video datasets capture first-person perspective footage from cameras worn or mounted on people, used by computer vision and embodied AI researchers to train models for hand-object interaction, activity recognition, and embodied task understanding. Researchers, roboticists, and vision model developers use these datasets to build systems that understand human actions and intentions from the wearer's viewpoint.
meta project aria · build ai · protege · defined.ai
Synthetic data and simulation generate digital environments, sensor data, and annotated scenarios for training AI models without collecting real-world footage. Robotics companies, autonomous vehicle developers, and computer vision teams use these systems to create labeled training datasets at scale.
nvidia isaac sim · nvidia cosmos · genesis ai · duality ai · parallel domain · bifrost · world labs · nvidia omniverse · duality robotics
Platforms that collect, organize, and manage sensor data (camera, lidar, radar) from vehicles for training autonomous driving systems. Used by AV companies, tier-1 suppliers, and research teams to build and validate perception models.
applied intuition · foretellix · segments.ai · kognic · deepen ai
Physical AI training data management involves collecting, organizing, and curating raw sensor data and video from robotic systems and physical environments for machine learning. Robotics companies, autonomous vehicle developers, and AI research labs use these platforms to store and version control training datasets.
foxglove · rerun · voxel51 · roboflow
Open robotics datasets are collections of robot sensor data, trajectories, and task demonstrations released publicly for training and benchmarking. Roboticists, ML researchers, and robot manufacturers use them to develop perception and control algorithms.
hugging face lerobot · physical intelligence · toyota research institute · google deepmind robotics · skild ai · nvidia gr00t · figure
Hardware kits that capture and collect raw training data for AI models through sensors, cameras, and recording devices. Used by robotics companies, research labs, and ML teams to gather real-world physical data for model training.
trossen robotics · manus · shadow robot
Physical AI training data annotation and labeling involves humans marking, categorizing, and structuring images, videos, and sensor data collected from physical environments to train machine learning models. Companies building autonomous systems, robotics platforms, and computer vision applications use these services to prepare raw collected data for model training.
labelbox · scale ai labeling · v7 · cvat · superb ai
Datasets of real-world driving scenes with sensor data (camera, lidar, radar) and annotations for training autonomous vehicle perception models. Used by automakers, robotics companies, and AI labs building self-driving systems.
waymo open dataset · nuscenes · mobileye · scale ai automotive