Myungkyu/RoboDojo-taco-visual2-gemini
RoboDojo-taco-visual2-gemini visual2 variant (2026-09-14): unlike RoboDojo-taco-visual-gemini (marker in a separate keyframe video), here the point marker (red disc, radius 1.9 % of the width, white ring) is drawn into the head camera video itself for every frame whose governing label names a position, and the label text says ... the location marked with the red dot ... (tic-tac-toe: ... <cell> marked with the red dot.). The dataset keeps the three live views only (no keyframe… See the full description on the dataset page: https://huggingface.co/datasets/Myungkyu/RoboDojo-taco-visual2-gemini.
dense-labelled episodes (TACOR release) part 10/10
dense-labelled episodes (TACOR release) part 9/10
dense-labelled episodes (TACOR release) part 8/10
dense-labelled episodes (TACOR release) part 7/10
dense-labelled episodes (TACOR release) part 6/10
dense-labelled episodes (TACOR release) part 5/10
dense-labelled episodes (TACOR release) part 4/10
dense-labelled episodes (TACOR release) part 3/10
dense-labelled episodes (TACOR release) part 2/10
dense-labelled episodes (TACOR release) part 1/10
initial commit