| Session Name: | Character Control with Neural Networks and Machine Learning |
| Speaker(s): | Daniel Holden |
| Company Name(s): | Ubisoft Montreal |
| Track / Format: | Programming |
| Overview: | Recent research has found that machine learning and neural networks can be used to construct animation systems for games in a revolutionary new way with far greater scalability and performance than was previously thought possible. At the same time, current methods for building AAA animation systems have started to suffer due to their sheer size and complexity. Using state of the art research into character animation and recent prototypes developed at Ubisoft's La Forge, this talk will show how many different kinds of data-driven systems can vastly reduce the complexity and man-power involved in building an animation system, while simultaneously allowing for a larger variety of high quality and interesting motion to be produced. As well as the numerous advantages, these approaches come with their own pitfalls and difficulties, many of which will also be covered in this practical and deep dive into machine learning for character animation. |
Recommended
