Open source and lab robotics dataset conversion and annotation
Converted approximately 1 million episodes of diverse robotic interaction datasets to a standardized model training format suitable for embodied AI research. Utilized Python programming to script data extraction, transformation, and conversion workflows for open-source and laboratory-collected episodes. Set up and operated physical lab equipment for the collection of robotic manipulation data and formatted this for downstream AI/ML modeling purposes. • Processed datasets from sources such as agibot-Beta, OXE, and robomind to Lerobot 2.1 format • Collected and formatted 600 data entries from physical robotic arm lab tasks • Developed data acquisition code using Python under Ubuntu and ROS2 environments • Ensured all processed data followed standardized structure for model training