Style-based inverse kinematics

By Keith Grochow, Steven L. Martin, Aaron Hertzmann, Zoran Popovic. ACM Transactions on Graphics, 2004

862 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Control and Systems Engineering

View Paper: https://www.semanticscholar.org/paper/69ab9e293ae3b54f7b004ac89d789716e0ea5aa4

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Control and Systems Engineering

Citations in the corpus

PiGraphs (2016, methodology): “Much prior work has addressed modeling of human motion style to allow for highlevel control and adaptation of motion data [Grochow et al. 2004; Shapiro 2011; Min and Chai 2012].”
Pose-conditioned joint angle limits for 3D human pose reconstruction (2015, background): “[10] proposed a scaled Gaussian latent variable model as a 3D pose prior.”
Gaussian Process Dynamical Models (2005, methodology): “For the latent mapping, , we currently use the RBF kerneíÜ Ü ¼ µ ¬ ½ ÜÔ ¬ ¾ ¾ Ü Ü ¼ ¾ · ¬ ½ ¿ AE ÜÜ ¼ (6) As in the SGPLVM [4], we use a scaling matrix Ï ´Û ½ Û µ to account for different variances in the different data dimensions.”
Style translation for human motion (2005, background): “Style-based inverse kinematics uses such an approach to infer a model of natural human poses [Grochow et al. 2004].”
People-tracking-by-detection and people-detection-by-tracking (2008, methodology): “This is equivalent to jointly solving the inverse kinematics in each frame of the sequence under soft constraints given by limb likelihoods and is similar to [9], except that in our case hints about limb positions are provided by a person detector instead of being manually given by the user.”
Local distance preservation in the GP-LVM through back constraints (2006, methodology): “Indeed the successful applications and extensions of the GP-LVM (Grochow et al., 2004; Urtasun et al., 2005; Shon et al., 2006; Wang et al., 2006) each take explicit advantage of the fact that the model is probabilistic.”
Using dimensionality reduction to exploit constraints in reinforcement learning (2010, background): “While this is a simple example, similar arguments also apply to more complicated situations, in particular for nat ur l movements with many degrees of freedom such as that of humans [10] or humanoid robots [11] subject to more complex environmental or task constraints [12].”
Dimension Reduction for Aerodynamic Design Optimization (2011, methodology): “Many variants of the method have been developed over recent years [45–48].”
Hierarchical Gaussian process latent variable models (2007, methodology): “…time is becoming a dominating cost in both the games and .lm entertainment industries where computer special e.ect techniques are used, through combination of the hierarchical GP-LVM with appropriate inverse kinematic techniques (Grochow et al., 2004) we could seek to ameliorate these costs.”
Vision-based human motion analysis: An overview (2007, background): “Reducing the dimensionality of the pose space is motivated by the observation that human activities are often located on a latent space that is low-dimensional [24,31].”
Robust motion in-betweening (2020, methodology): “Other statistical models, such as Gaussian Processes [Min and Chai 2012] and Gaussian Process Latent Variable Models [Grochow et al. 2004; Levine et al. 2012; Wang et al. 2008; Ye and Liu 2010] have been applied to the constrained motion control task, but are often limited by heavy run-time…”
Mode-adaptive neural networks for quadruped motion control (2018, background): “…Safonova et al. 2004; Tautges et al. 2011], radial basis functions (RBF) [Kovar and Gleicher 2004; Rose et al. 1998], reinforcement learning [Safonova and Hodgins 2007] and Gaussian processes (GP) [Grochow et al. 2004; Ikemoto et al. 2009; Mukai and Kuriyama 2005] are studies to achieve this goal.”
Phase-functioned neural networks for character control (2017, methodology): “Gaussian Process Latent Variable Models (GPLVM) are applied for computing a low dimensional latent space of the data to help solve the redundancy of inverse kinematics problems [Grochow et al. 2004] or to improve the efficiency for planning movements [Levine et al. 2012].”