← ArchivePaper2004

Style-Based Inverse Kinematics

Keith Grochow, Steven L. Martin, Aaron Hertzmann, Zoran Popovic

SIGGRAPHAcademic2 descendantsRigging Retargeting

Gaussian process model for style-aware inverse kinematics generating human-like poses respecting learned stylistic priors.

Open the paper ↗Trace in the lineage graph

Abstract

This paper presents an inverse kinematics system based on a learned model of human poses. Given a set of constraints, our system can produce the most likely pose satisfying those constraints, in real-time. Training the model on different input data leads to different styles of IK. The model is represented as a probability distribution over the space of all possible poses. This means that our IK system can generate any pose, but prefers poses that are most similar to the space of poses in the training data. We represent the probability with a novel model called a Scaled Gaussian Process Latent Variable Model. The parameters of the model are all learned automatically; no manual tuning is required for the learning component of the system. We additionally describe a novel procedure for interpolating between styles.Our style-based IK can replace conventional IK, wherever it is used in computer animation and computer vision. We demonstrate our system in the context of a number of applications: interactive character posing, trajectory keyframing, real-time motion capture with missing markers, and posing from a 2D image.

How to read this

Category: Method: learned-prior inverse kinematics (probabilistic pose model)
Contributions: An IK system that, given constraints, produces the most likely pose in real time under a learned probability distribution over poses
A novel Scaled Gaussian Process Latent Variable Model whose parameters are learned automatically with no manual tuning, where training data determines the IK style
A procedure for interpolating between styles, demonstrated on posing, trajectory keyframing, real-time mocap with missing markers, and posing from a 2D image
Context: Sits in the data-driven, statistical character-posing lineage, applying Gaussian-process latent-variable modeling as a learned pose prior to constrain inverse kinematics toward human-like results.
Correctness: The system can reach any pose but is biased toward the training distribution, so style and plausibility depend on the training data; readers should note out-of-distribution constraints will be pulled toward learned poses, which is the intended behavior but a limitation for novel motions.
Clarity: The application framing is accessible, but the SGPLVM core is mathematically dense; a first pass conveys the learned-prior idea, a second and likely third pass are needed for the model.
How to read it: Read first for the probabilistic-IK framing and the demos that show what the prior buys you; budget a careful second/third pass on the Scaled GPLVM if you need the formulation.

Builds on

Nothing in the archive, this is a starting point.

Built upon by

Related work

Vision-based control of 3D facial animation 2003 / SCA
A Facial Motion Retargeting Pipeline for Appearance Agnostic 3D Characters 2024 / CASA
Optimal and Interactive Keyframe Selection for Motion Capture 2018 / SIGGRAPH Asia
Creating an Actor-Specific Facial Rig from Performance Capture 2016 / DigiPro

Keywords

style-based inverse kinematics scaled gaussian process latent variable model gaussian process latent variable model probability density function character posing motion capture missing markers non-linear dimensionality reduction style interpolation kernel function active set trajectory keyframing pose synthesis maximum likelihood 2d image posing

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 →