Full Body Tracking from Multiple Views Using Stochastic Sampling

Kehl, Roland; Bray, Matthieu; Gool, Luc Van

doi:10.1109/cvpr.2005.165

Cited by 88 publications

(60 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The model-based (or generative) approaches [10,12,19,17] are usually expressed within the analysis-by-synthesis paradigm. An explicit model is usually designed which is similar to the target (observation) and an error measure between these two is defined and then minimized at each frame.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Regression-Based Human Motion Capture From Voxel Data

Sun¹,

Bray²,

Thayananthan³

et al. 2006

Procedings of the British Machine Vision Conference 2006

Self Cite

View full text Add to dashboard Cite

A regression based method is proposed to recover human body pose from 3D voxel data. In order to do this we need to convert the voxel data into a feature vector. This is done using a Bayesian approach based on Mixture of Probabilistic PCA that transforms a collection of 3D shape context descriptors, extracted from the voxels, to a compact feature vector. For the regression, the newly-proposed Multi-Variate Relevance Vector Machine is explored to learn a single mapping from this feature vector to a low-dimensional representation of full body pose. We demonstrate the effectiveness and robustness of our method with experiments on both synthetic data and real sequences.

show abstract

Section: Introductionmentioning

confidence: 99%

“…Our method belongs to regression-based sub-category and then differs from other model-based methods which fit a body model to the voxel data (e.g. [10,12]). …”

Section: Introductionmentioning

confidence: 99%

Regression-Based Human Motion Capture From Voxel Data

Sun¹,

Bray²,

Thayananthan³

et al. 2006

Procedings of the British Machine Vision Conference 2006

Self Cite

View full text Add to dashboard Cite

show abstract

“…Kehl et al [4] also proposed a markerless solution for full-body pose tracking.A model built from super-ellipsoids is fitted to a colored volumetric reconstruction using stochastic meta descent (SMD), taking advantage of the color information to overcome ambiguities caused by limbs touching each other. To increase robustness and accuracy, the tracking is refined by matching model contours against image edges.…”

Section: Trackingmentioning

confidence: 99%

“…Expensive projections of voxels can be avoided and the algorithm can take advantage of small changes in the images by addressing only voxels whose pixel has changed. around, as proposed by Kehl et al [4]. Instead of projecting the voxels into the camera views at each frame, we keep a fixed lookup table (LUT) for each one and store a list at each pixel with pointers to all voxels that project onto that particular pixel (see Figure 14.6).…”

Section: Reconstructionmentioning

confidence: 99%

“…Then the segmentations from the different cameras are combined to make CHAPTER 14 Real-Time 3D Body Pose Estimation a 3D hull reconstruction. This is done in real time using voxel carving and a fixed lookup table [4]. The camera network is distributed, as each camera is connected to a separate PC that runs the foreground-background segmentations.The segmentations are sent to a central PC that runs the hull reconstruction.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Real-Time 3D Body Pose Estimation

Bergh¹,

Koller-Meier²,

Kehl³

et al. 2009

Multi-Camera Networks

View full text Add to dashboard Cite

Feature-Based Pose Estimation

Sminchisescu

Ionescu

et al. 2011

Visual Analysis of Humans

View full text Add to dashboard Cite

In this chapter we review challenges and methodology for feature-based predictive tri-dimensional human pose reconstruction, based on image and video data. We argue that reliable 3d human pose prediction can be achieved through an alliance between image descriptors that encode multiple levels of selectivity and invariance and models that are capable to represent multiple structured solutions. For monocular systems, key to reliability is the capacity to leverage prior knowledge in order to to bias solutions not only to kinematically feasible sets, but also towards typical configurations that humans are likely to assume in everyday surroundings. In this context, we discuss several predictive methods including large-scale mixture of experts, supervised spectral latent variable models and structural support vector machines, asses the impact of the various choices of image descriptors, review open problems, and give pointers to datasets and code available online.

show abstract

Full Body Tracking from Multiple Views Using Stochastic Sampling

Cited by 88 publications

References 12 publications

Regression-Based Human Motion Capture From Voxel Data

Regression-Based Human Motion Capture From Voxel Data

Real-Time 3D Body Pose Estimation

Feature-Based Pose Estimation

Contact Info

Product

Resources

About