Open-H-Embodiment: A Large-Scale Dataset for Enabling Foundation Models in Medical Robotics

The Open-H-Embodiment Consortium, a global collaboration of over 212 researchers from institutions like Johns Hopkins, UC Berkeley, and others, has released the largest open dataset of medical robotic video with synchronized kinematics. This dataset is designed to overcome the long-standing data bottleneck in autonomous medical robotics, where existing datasets are typically small, limited to a single robot embodiment, and rarely shared openly. By providing a diverse collection of surgical demonstrations across multiple robot platforms, including the da Vinci system, the dataset enables the training of foundation models that can generalize across different tasks and robots, a critical step toward autonomous surgical assistance.

Open-H-Embodiment includes thousands of hours of video paired with precise kinematic data, capturing a wide range of surgical procedures such as suturing, knot-tying, and tissue manipulation. The dataset's scale and diversity are unprecedented in the field, offering a rich resource for developing AI models that can learn from varied demonstrations and adapt to new scenarios. This initiative aims to accelerate research into autonomous medical robots, which promise to improve patient outcomes by reducing human error, democratizing access to expert-level surgical care, and enabling superhuman precision in complex procedures. The dataset is publicly available to foster collaboration and innovation across the medical robotics community.