Case Studies
Real results from robotics teams who trained on Radioception datasets.
Challenge
Needed 1,200+ demonstrations of mixed-weight box palletizing to train a new grasping policy. Off-the-shelf datasets lacked force-torque data and their proprietary end-effector geometry.
Challenge
Building a robot-assisted suture training system. Required sub-millimeter precision keypoint annotations of expert surgeon hand movements — impossible to source from any existing dataset.
Challenge
Training a home cleaning robot to handle kitchen tasks across diverse object types. Needed 2,000+ pouring demonstrations with varying liquid viscosities, vessel sizes, and lighting conditions.
Challenge
Training a 7-DoF arm for peg-in-hole assembly tasks with tolerances under 0.5mm. Existing datasets lacked the required force feedback annotations for high-precision manipulation.
Also working with