RoMo-HML-272: Human Body Motion in 272-Dimensional Features

RoMo-HML-272 encodes body motion using a 272-dimensional feature representation. It augments the standard 263-D HumanML3D encoding with 9 additional dimensions for absolute-3D and contact-augmented features. The dataset is authored by RoMoDataset and is scheduled for public release alongside a forthcoming paper.

Use Cases

• Train motion generation models based on the 272-dimensional HumanML3D-style feature representation.
• Benchmark novel motion synthesis techniques using the described absolute-3D and contact-augmented features.
• Analyze human body motion patterns from the structured, high-dimensional feature vectors.

Strengths

• Features are structured in a 272-dimensional vector, building on a known standard.
• Augments the 263-D HumanML3D encoding with 9 specific dimensions for absolute-3D and contact features.

Limitations

• Dataset is currently private; access is restricted until the associated paper is released.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

RoMoDataset

Freshness

Last updated 2026-05-21 18:24:20; freshness should be verified.