Real-time Animation of Dressed Virtual Humans

Cloth simulation has a long history, starting almost three decades ago. Thanks to the continuous increase of computation speed of computers, real-time simulation of clothes has become very recently possible. The research on real-time clothes has really begun few years ago when researchers have started simulating simple garments such as flags or tablecloths. Nowadays, the research aims at proposing methods that can simulate cloth in real-time independently of its complexity. Most of the existing approaches for real-time cloth simulation consider clothes as independent object and are able to simulate the dynamics of cloth either as clothing or other garments like in furnishing tablecloth. Such assumption requires implementing a mass spring system with collision detection. The main drawback of these approaches is that they are slow at computation, making the simulation of dressed humans hardly achievable. In this dissertation, we propose to consider clothes as being anchored to the underlying skin surface rather than as independent object. With such an assumption, many optimizations can be made. Collision detection can be restricted to the cloth vertices and their anchorage on the skin. Cloth dynamics can be simulated with a faster method than the usual particle system. Starting from this idea, two different approaches have been developed and experimented. The first approach works in a hybrid manner exploiting the merits of both the physically based and geometric deformations. It makes use of predetermined conditions between the cloth and the body model, avoiding complex collision detection and physical deformations wherever possible. Garments are segmented into pieces that are simulated by various algorithms, depending on how they are laid on the body surface and whether they stick or flow on it. Tests show that the method is well suited to fully dressed virtual human models, achieving real-time performance compared to ordinary cloth-simulations. The main idea of the second approach consists of developing a cloth simulator that can learn the cloth behaviour through a sequence of pre-computed cloth simulation. This approach enables us to simulate the garment features such as wrinkles, pockets... in real-time. We setup the problem as simulating the garments in two phases. The first phase, a rough mesh reproduces the dynamic behaviour of the garments. Its physical properties are defined by the pre-calculated sequence. In the second phase, the fine mesh simulates the details of the garments. Finally, we introduce one application of the real-time clothes: the Virtual Try-On. This application aims to provide to customers a mannequin of their size wearing the cloth of their choice.