← ArchivePaper2019
Learning an Intrinsic Garment Space for Interactive Authoring of Garment Animation
Encodes garment deformations in a low-dimensional intrinsic space learned from simulation, enabling interactive authoring and blending of garment animations.
Abstract
Authoring dynamic garment shapes for character animation on body motion is one of the fundamental steps in the CG industry. Established workflows are either time and labor consuming (i.e., manual editing on dense frames with controllers), or lack keyframe-level control (i.e., physically-based simulation). Not surprisingly, garment authoring remains a bottleneck in many production pipelines. Instead, we present a deep-learning-based approach for semi-automatic authoring of garment animation, wherein the user provides the desired garment shape in a selection of keyframes, while our system infers a latent representation for its motion-independent intrinsic parameters (e.g., gravity, cloth materials, etc.). Given new character motions, the latent representation allows to automatically generate a plausible garment animation at interactive rates. Having factored out character motion, the learned intrinsic garment space enables smooth transition between keyframes on a new motion sequence. Technically, we learn an intrinsic garment space with an motion-driven autoencoder network, where the encoder maps the garment shapes to the intrinsic space under the condition of body motions, while the decoder acts as a differentiable simulator to generate garment shapes according to changes in character body motion and intrinsic parameters.
How to read this
- Category
- Method: learning-based garment animation authoring
- Contributions
- Learns an intrinsic garment space with a motion-driven autoencoder, factoring out body motion from motion-independent parameters such as gravity and cloth material
- Lets artists specify garment shape on a few keyframes and infers a latent intrinsic representation for the rest
- Generates plausible garment animation on new character motion at interactive rates with smooth keyframe-to-keyframe transitions
- Context
- Addresses the keyframe-control-versus-simulation tradeoff in garment authoring, in the same learning-based cloth lineage as Santesteban et al.'s 'Learning-Based Animation of Clothing for Virtual Try-On' (2019).Builds on: Learning-Based Animation of Clothing for Virtual Try-On
- Correctness
- The method assumes garment behavior can be disentangled into body-motion-conditioned and motion-independent intrinsic factors learned from simulation; this enables interactive authoring but ties output quality and plausibility to the coverage and realism of the training simulations.
- Clarity
- Accessible framing of the authoring problem; a first pass conveys the keyframe-plus-latent idea, with the encoder conditioning details warranting a second pass.
- How to read it
- Read first for the authoring workflow and the motion-conditioned autoencoder concept; revisit the network design and the intrinsic-versus-motion factorization if you care about controllability or reproducing the system.
Builds on
Built upon by
Nothing yet.
Related work
- PBNS: Physically Based Neural Simulation for Unsupervised Garment Pose Space Deformation 2021 / SIGGRAPH Asia
- NeuroSkinning: Automatic Skin Binding for Production Characters with Deep Graph Networks 2019 / SIGGRAPH
- Physics-Inspired Upsampling for Cloth Simulation in Games 2021 / SIGGRAPH
- Strain Based Dynamics 2014 / SCA
Keywords
This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →