Unique Instances for Crowds

The simulation of large crowds in real time requires many instances of similar characters. We present an algorithm to allow for each individual in the crowd to be unique. Our contribution is twofold: first, a simple and efficient method for attaching accessories to individuals so as to modify their look. Second, a new and generic technique based on segmentation maps to add detailed color variety and patterns to human meshes as well as accessories. Both methods are scalable to suit all human levels of detail exploited in crowd simulations, i.e., impostors, rigid and deformable meshes. Tests and comparisons show that while offering an appealing visual aspect to the crowd, the presented approach is adapted for real-time simulations of thousands of unique characters.

Category:

Science & Technology

Tags:

• unique
• virtual
• crowds
• simulation
• variety

[theogore]

Stunning work.

Too bad this couldn't be an actual app to make machinima with...