[Open-source] Motion Control with Neural Networks Demo

Hey guys I think some of you may have seen this paper a few weeks ago:

This demo was implemented in Unity and the authors (Zhang and Stark) have made it open-source at: GitHub - sebastianstarke/AI4Animation: Bringing Characters to Life with Computer Brains in Unity

I’ve created a short video explaining how this method works and a tutorial on how you can use it yourself with your data in your games. Here is the link for it: http://3deeplearner.com/siggraph-02

Although the authors focus on quadrupeds this really works with any data you throw at. Here is a demo a did with data from the CMU mocap dataset:

I hope this is the most appropriate place to post this.
Cheers,
Gus

2 Likes

this is absolutely fabulous.

do you think this could work with other type of movements, like walking/running backwards, strafing etc?

Thanks.
Yes there is nothing significantly different about this movements. They should be possible, provided you have the training data and you set the appropriate styles in the character controller.