Datasets:
A few false-start episodes (20, 37, 38, 39, 55)
#2
by maxmccutcheon59 - opened
Hi LeRobot team, thanks for these datasets! An open-source quality checker I've been building flagged a few episodes here: episode 20 is a single frame, and in episodes 37, 38, 39 and 55 the commanded actions never change (the robot doesn't move); 37 and 38 are also under a second long. They look like accidental recordings and might be worth excluding.
Details: github.com/maxmccutcheon59/robolint/blob/main/AUDIT.md
Reproduce with: robolint check lerobot/svla_so100_stacking