RoMo-272 is the RoMo body-motion corpus (paper core) packed in the 272-dimensional motion representation of Li-xingXiao/272-dim-Motion-Representation — a HumanML3D-derived encoding that augments the standard 263-D HumanML3D layout with 9 additional absolute/global features used by several recent text-to-motion methods. Each clip carries five text captions and a three-level semantic taxonomy, with fixed train / val /… See the full description on the dataset page:
https://huggingface.co/datasets/RoMoDataset/RoMo-272.
