Algorithms for deep motion generation

With the recent development of learning based methods, many works have focused on generative models for motion synthesis. Most learning-based methods treat joint motions as the prediction target while considering their spatial and temporal connections. In our library, we include a wide range of deep-learning methods from RNN, VRNN,`VAE`, GAN to Transformer that train generative models on different datasets:

• On human motion prediction using recurrent neural networks
• Action-Conditioned 3D Human Motion Synthesis with Transformer VAE
• A Recurrent Variational Autoencoder for Human Motion Synthesis
• Action2Motion: Conditioned Generation of 3D Human Motions