QuaterNet: A Quaternion-Based Recurrent Model for Human Motion

This is the implementation of the approach described in the paper:

We provide the code for reproducing our results (short-term prediction) and generating/rendering locomotion animations (long-term generation), as well as pre-trained models. Our recurrent network, QuaterNet, represents rotations with quaternions and our loss function performs forward kinematics on a skeleton to penalize absolute position errors instead of angle errors. If you want to train the model from scratch, you have to run the download script first:

This will download the Human3.6M dataset, convert it to our encoding, and save it in .

Source: github.com