AnthroNet: Conditional Generation of Humans via Anthropometrics

• DOI: 10.48550/arxiv.2309.03812
• Publication Date: 2023-01-01
• Authors: Francesco Picetti; Shrinath Deshpande; Jonathan Leban; Soroosh Shahtalebi; Jay Patel; Peifeng Jing; Chunpu Wang; Charles Metze; Cameron Sun; Cera Laidlaw; James Warren; Kathy Huynh; River Page; Jonathan Hogins; Adam Crespi; S. Ganguly; Salehe Erfanian E...
• Keywords: Human body

We present a novel human body model formulated by an extensive set of anthropocentric measurements, which is capable generating wide range shapes and poses. The proposed enables direct modeling specific identities through deep generative architecture, can produce humans in any arbitrary pose. It the first its kind to have been trained end-to-end using only synthetically generated data, not provides highly accurate mesh representations but also allows for precise anthropometry body. Moreover, diverse animation library, we articulated our synthetic humans' hands maximize diversity learnable priors training. Our was on dataset $100k$ procedurally-generated posed meshes their corresponding anthropometric measurements. data generator be used generate millions unique poses non-commercial academic research purposes.