(2) Deep Neural Networks for Motion Modeling and Synthesis:
The idea for this practical exercise session is to give access to our motion modelling and synthesis pipeline that makes use of various machine learning techniques such as convolutional neural networks and PCA. In practical terms, the participants in this exercise session would capture motions, semantically annotate these motions in a semi-automated manner, possibly cut motions into motion primitives, and produce motion models for these motion primitives. Depending on whether the approach chosen by the participants allows for it, the participants can also decorate the motion models with different motion styles (male, female, old, child, bold etc.). The resulting motions can then be visualized in a predefined environment.
The number of participants for this practical exercise is limited to 24 (max. 8 groups of 3) due to limited availability of GPUs. No physical computer but only remote access to the GPU will be provided: Participants need to bring their own laptop
Tutors: Klaus Fischer, Janis Sprenger, Somayeh Hosseini, Noshaba Cheema, Han Du, Erik Herrmann
|