Retarget Quality Evaluation

User study for G1 humanoid retargeting quality (v3)

📋 Task:

For each sequence, compare 4 retarget methods and rank them from best to worst using 1, 2, 3, 4 (1 = best, 4 = worst). Ranks must be a permutation — no ties.

Rank along these 5 criteria:

⚠ When the original human motion (the reference at the top) is noisy/jittery, a result that refines it (cleaner, less jittery) is preferred even if it deviates from the noisy original.

Differences between methods can be subtle — please watch closely and rank carefully. Use the playback controls (space = play/pause, ← / → = frame step) to inspect frame-by-frame.

Enter your name to start.
(admin-amaya / admin-yuto route to the admin dashboard)


Past experiments (archive)