Overview
Save Big on Coursera Plus. 7,000+ courses at $160 off. Limited Time Only!
Explore how data-driven systems can revolutionize character animation in game development through this 2018 GDC talk by Ubisoft's Daniel Holden. Discover techniques for reducing complexity and manpower in building animation systems for character control using neural networks and machine learning. Delve into topics such as data separation, fuzzy matching, generalization, and the fundamentals of neural networks. Learn about practical applications, including quadruped training, overlays, and incremental training. Gain insights into switching between different systems, fine-tuning, and layering techniques. Understand the process of creating and managing training data for neural network-based animation systems.
Syllabus
Intro
Background Animation System
The Problem
Data Separation
Specifying desired variables
Fuzzy matching
Generalizing
Example
How Neural Networks Work
Bloopers
Questions
Tagging
Incremental Training
Quadruped Training
Overlays
Training Data
Switching to Different Systems
Fine Tuning
Layering
Neural Network Training
Taught by
GDC