PBNS

We present a methodology to automatically obtain Pose Space Deformation (PSD) basis for rigged garments through deep learning. Classical approaches rely on Physically Based Simulations (PBS) animate clothes. These are general solutions that, given sufficiently fine-grained discretization of space and time, can achieve highly realistic results. However, they computationally expensive any scene modification prompts the need re-simulation. Linear Blend Skinning (LBS) with PSD offers lightweight alternative PBS, though, it needs huge volumes data learn proper PSD. propose using learning, formulated as an implicit un-supervisedly cloth Deformations in constrained scenario: dressed humans. Furthermore, we show is possible train these models amount time comparable PBS few sequences. To best our knowledge, first neural simulator cloth. While deep-based domain becoming trend, data-hungry models. Moreover, authors often complex formulations better wrinkles from data. Supervised learning leads physically inconsistent predictions that require collision solving be used. Also, dependency limits scalability solutions, while their formulation hinders its applicability compatibility. By proposing unsupervised LBS (3D animation standard), overcome both drawbacks. Results obtained cloth-consistency animated meaningful pose-dependant folds wrinkles. Our solution extremely efficient, handles multiple layers cloth, allows outfit resizing easily applied custom 3D avatar.