Encord released EMG-1, an open‑source collection of more than 300 hours of synchronized muscle‑signal (EMG) and first‑person video data. The dataset is about 30 times larger than any similar public set. Tests showed that adding EMG to video lowered fingertip tracking error by 25 % and reduced errors during short visual gaps by 35 %. Researchers can download the data from encord.com/data‑labs.
Why it matters
Robotics teams can improve hand‑pose tracking in occluded situations, leading to more reliable manipulation in industrial robots.
