Stochastic Tracking of 3D Human Figures Using 2D Image Motion

By Hedvig Kjellström, Michael J. Black, David J. Fleet. European Conference on Computer Vision, 2000

866 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/af979bb313d15d55f1956f3929675a607b200e3b

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

Citations in the corpus

Finding and tracking people from the bottom up (2003, background): “Particle filtering uses multiple predictions — obtained by running samples of the prior through a model of the dynamics — which are refined by comparing them with the local image data (the likelihood) (see, for example [14, 3]).”
Implicit Probabilistic Models of Human Motion for Synthesis and Tracking (2002, methodology): “Here we use a database-specific basis set learned using PCA.”
On Space-Time Interest Points (2005, neutral): “Some of the works rely on pure spatial image features while using sophisticated body models and tracking schemes to constrain the interpretation (Baumberg and Hogg, 1996; Bregler and Malik, 1998; Sidenbladh, Black and Fleet, 2000).”
A Multiview Approach to Tracking People in Crowded Scenes Using a Planar Homography Constraint (2006, background): “These and other similar algorithms [6, 7, 8] are challenged by occluding and partially occluding objects, as well as appearance changes.”
MoDeep: A Deep Learning Framework Using Motion Features for Human Pose Estimation (2014, background): “Several other model based articulated tracking systems have been reported over the past two decades, most notably [10–16].”
Strike a pose: tracking people by finding stylized poses (2005, positive): “Particle filtering uses multiple predictions – obtained by running samples of the prior through a model of the dynamics – which are refined by comparing them with the local image data (the likelihood) (see, for example [2] [9] [19]).”
A survey on visual surveillance of object motion and behaviors (2004, methodology): “These motion models serve as prior knowledge to predict motion parameters [106], [107], to interpret and recognize human behaviors [108], or to constrain the estimation of low-level image measurements [109].”
Kinematic jump processes for monocular 3D human tracking (2003, methodology): “Sidenbladhet al use particle filtering with importance sampling based on either a learned walki g model or a database of motion snippets, to focus search in the neighborhood of known trajectory pathways [11, 12].”
Video-based event recognition: activity representation and probabilistic recognition methods (2004, methodology): “Robust tracking often requires an object model and a sophisticated optimization process [22].”
Tracking loose-limbed people (2004, background): “5D [17], or 3D [2, 5, 21, 23] leading to a high-dimensional parameter space (25–50 dimensions is not uncommon).”
3D Pictorial Structures for Multiple Human Pose Estimation (2014, background): “The problem is often coupled with tracking [7, 9, 13, 21, 28, 30].”
Tracking multiple humans in complex situations (2004, methodology): “Motion subspace is used in Sidenbladh et al. [37] to track human walking using a particle filter.”
Hand gesture recognition using multi-scale colour features, hierarchical models and particle filtering (2002, methodology): “Our implementation of particle filtering largely follows the traditional approaches for condensation as presented in [8, 1, 18].”
HumanEva: Synchronized Video and Motion Capture Dataset and Baseline Algorithm for Evaluation of Articulated Human Motion (2010, methodology): “While no “standard” algorithm exists in the community, we implemented a fairly common Bayesian filtering method based on the methods of Deutscher and Reid (2005) and Sidenbladh et al. (2002).”
Dynamical binary latent variable models for 3D human pose tracking (2010, background): “Early dynamical models were formulated as smooth, linear Markov models [2] [3] [18], but such models do not capture nonlinear properties of human motion and were found insufficient for monocular 3D pose tracking.”
Human3.6M: Large Scale Datasets and Predictive Methods for 3D Human Sensing in Natural Environments (2014, background): “matic discriminative prediction [10], [11], [12], [13], [14], whereas others aim at model-image alignment [15], [16], [17], [18], [19], [20], [21], [22], [23] or accurate modeling of 3D shape[24], [25] or clothing[26].”
3D hand tracking by rapid stochastic gradient descent using a skinning model (2004, background): “[1] manage to reduce the number of parameters in the state vector.”
Vision-based human motion analysis: An overview (2007, methodology): “Regarding the number of cameras that is used, monocular work [38,3,105,93] is appealing since for many applications only a single camera is available.”