HOOD: Real-Time Prediction of Realistic Clothing Dynamics

Researchers have presented a new method for real-time prediction of realistic clothing dynamics using graph neural networks, multi-level message passing, and unsupervised training. 

The method, called HOOD, takes into consideration body shapes and applies to any kind of clothing  It also handles changes in topology and material properties at inference time. 

The researchers proposed a "hierarchical message-passing scheme that efficiently propagates stiff stretching modes while preserving local detail" and say that their method provides more realistic results than state-of-the-art ones. 

If you're interested in the technical side of the project, check out the paper here. Also, don't forget to join our 80 Level Talent platform and our Telegram channel, follow us on Instagram and Twitter, where we share breakdowns, the latest news, awesome artworks, and more.