The RANDOM People dataset is a collection of 188 crowd-sourced videos featuring diverse human identities performing specific motion sequences designed for 3D Gaussian avatar creation. This dataset was developed to support synthetic human action video generation research, enabling the creation of photorealistic videos where novel identities can reenact human actions from reference videos. The dataset addresses the critical need for diverse human representation in computer vision datasets… See the full description on the dataset page:
https://huggingface.co/datasets/matybohacek/RANDOM-People.
