Hokyun Im

Upscaled dataset (256 by 256). This includes filtered dataset (no_noops) for BC, but also original dataset for RL. Followed OpenVLA's postprocessing.