LogInThe model allows for realistic and accurate control of 3D full-body avatars in AR/VR applications.
The Meta AI Research team has recently presented Avatars Grow Legs (AGRoL), a novel conditional diffusion model designed to track full bodies with sparse upper-body tracking signals. The model employs a multi-layer perceptrons (MLP) architecture and a novel conditioning scheme for motion data, resulting in accurate and smooth full-body motion prediction, especially for lower body movement.
According to the team, AGRoL is capable of real-time processing, making it well-suited for online body-tracking applications. The model was trained on the AMASS motion capture dataset and, as stated by the developers, has demonstrated superior performance compared to state-of-the-art methods in terms of motion accuracy and smoothness.
"Our network is composed of only 4 types of components widely used in the deep learning era: fully connected layer (FC), SiLU activation layer, 1D convolutional layer with kernel size 1 and layer normalization(LN)," reads the team's research paper. "Note that the 1D convolutional layer with kernel size 1 can be also seen as a fully connected layer operating on a different dimension. Initialized from gaussian noises, the motion sequence is fed to the network after combining with the upper body signals."
You can learn more here and access the code over 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.