Supplementary materials

RoMo Hands: A Large Scale Richly Organized Text‑to‑Hand Motion Dataset

Interactive supplementary artifacts for the RoMo Hands dataset paper. Explore aggregate statistics, browse representative motion sequences in 3D, and compare ground-truth captures against generative baselines.

TL;DR Five interactive supplementary artifacts for the RoMo Hands paper. Start with Statistics for scale, Explorer / Grid for qualitative diversity, Comparisons for visual baseline behavior, and Benchmark for the numbers.

Browse

Dataset Statistics

Interactive dashboard summarizing sequence counts, durations, activity levels, and per-finger dynamics across RoMo Hands and the comparison datasets.

Open dashboard

Hand Motion Explorer

Searchable, filterable card-grid viewer over a 20-clip subset of RoMo Hands. Each card opens a full MANO viewer with timeline scrubber, five-level captions, taxonomy labels and per-clip quality scores.

Open explorer

Synchronized Motion Grid

Single-scene 3D viewer rendering all 20 representative bimanual MANO sequences in a synchronized grid. Click a tile to focus, scrub and inspect one motion at a time.

Open grid viewer

Generative Method Comparisons

Side-by-side MANO viewers comparing ground-truth motion against generative baselines (MDM, MoMask, MotionMillion, Kimodo) across five caption levels.

Open comparison

Generative Method Benchmark

Quantitative evaluation of generative baselines: taxonomy-aware semantic retrieval (R@K, MM-Dist, CLIP), distributional metrics (FID, diversity), and physical plausibility.

Open benchmark