Learning Generative Models for Monocular Body Pose Estimation

Jaeggli, Tobias; Koller-Meier, Esther; Gool, Luc Van

doi:10.1007/978-3-540-76386-4_57

Cited by 11 publications

(12 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For physics-based models with dynamics, related works include [112,113]. When temporal information is available, prior models [109] of human motion can be learned to constrain the inference of 3D pose sequences to improve monocular human pose tracking.…”

Section: Human Body Modelsmentioning

confidence: 99%

Human Pose Estimation from Monocular Images: A Comprehensive Survey

Gong

Zhang

Gonzàlez

et al. 2016

Sensors

124

View full text Add to dashboard Cite

Human pose estimation refers to the estimation of the location of body parts and how they are connected in an image. Human pose estimation from monocular images has wide applications (e.g., image indexing). Several surveys on human pose estimation can be found in the literature, but they focus on a certain category; for example, model-based approaches or human motion analysis, etc. As far as we know, an overall review of this problem domain has yet to be provided. Furthermore, recent advancements based on deep learning have brought novel algorithms for this problem. In this paper, a comprehensive survey of human pose estimation from monocular images is carried out including milestone works and recent advancements. Based on one standard pipeline for the solution of computer vision problems, this survey splits the problem into several modules: feature extraction and description, human body models, and modeling methods. Problem modeling methods are approached based on two means of categorization in this survey. One way to categorize includes top-down and bottom-up methods, and another way includes generative and discriminative methods. Considering the fact that one direct application of human pose estimation is to provide initialization for automatic video surveillance, there are additional sections for motion-related methods in all modules: motion features, motion models, and motion-based methods. Finally, the paper also collects 26 publicly available data sets for validation and provides error measurement methods that are frequently used.

show abstract

Section: Human Body Modelsmentioning

confidence: 99%

Human Pose Estimation from Monocular Images: A Comprehensive Survey

Gong

Zhang

Gonzàlez

et al. 2016

Sensors

124

View full text Add to dashboard Cite

show abstract

“…While existing methods typically treat gestures as single commands, such as "click the mouse", the proposed gesture manifold model enables us to automatically recognize the performed gesture and to track the movements within a gesture for fine-tuning continuous parameters. Manifold learning techniques have shown to provide compact, low-dimensional representations of human motion data [11] and have been used for human pose tracking [12,8]. We combine multiple, gesture-specific manifold models and subdivide the embeddings into phases allowing us to assign particular poses to arbitrary parameter settings.…”

Section: Introductionmentioning

confidence: 99%

Learning Gestures for Customizable Human-Computer Interaction in the Operating Room

Schwarz

Bigdelou

Navab

2011

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. Interaction with computer-based medical devices in the operating room is often challenging for surgeons due to sterility requirements and the complexity of interventional procedures. Typical solutions, such as delegating the interaction task to an assistant, can be inefficient. We propose a method for gesture-based interaction in the operating room that surgeons can customize to personal requirements and interventional workflow. Given training examples for each desired gesture, our system learns low-dimensional manifold models that enable recognizing gestures and tracking particular poses for fine-grained control. By capturing the surgeon's movements with a few wireless body-worn inertial sensors, we avoid issues of camera-based systems, such as sensitivity to illumination and occlusions. Using a component-based framework implementation, our method can easily be connected to different medical devices. Our experiments show that the approach is able to robustly recognize learned gestures and to distinguish these from other movements.

show abstract

“…Prior models have been used for constraining the tracking problem from observations such as silhouettes in monocular videos [10,23] or wearable sensor data [20]. The common idea is to avoid searching the high-dimensional full-body pose space and to use a learned parametrisation of feasible human poses instead.…”

Section: Related Workmentioning

confidence: 99%

“…Several authors have proposed tracking methods based on the Gaussian Process Latent Variable Model (GPLVM) [22,23], where a low-dimensional latent space of poses for a given activity is learned from training data. Other authors use manifold learning techniques, such as Isomap [10] or Laplacian Eigenmaps [17], for obtaining low-dimensional pose priors. We also choose a manifold learning method over GPLVMs since the latter are computationally significantly more expensive.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Manifold Learning for ToF-based Human Body Tracking and Activity Recognition

Schwarz¹,

Mateus²,

Castañeda³

et al. 2010

Procedings of the British Machine Vision Conference 2010

View full text Add to dashboard Cite

In this paper, we propose a method for simultaneous human full-body pose tracking and activity recognition from time-of-flight (ToF) camera images. Simple and sparse depth cues are used together with a prior motion model that constrains the tracking problem. Our model consists of low-dimensional manifolds of feasible poses for multiple activities. A particle filter allows us to efficiently evaluate various pose hypotheses over different activities and to select one that is most consistent with the observed depth image cues. We relate poses in the manifold embeddings to full-body poses and to observable depth cues using non-linear regression mappings. Our method is able to robustly detect changes of activity and adapt accordingly. We evaluate our method on a dataset containing 10 activities for 10 persons and show that we can track full-body pose and classify performed activities with a high precision which is discussed in the paper.

show abstract

Learning Generative Models for Monocular Body Pose Estimation

Cited by 11 publications

References 10 publications

Human Pose Estimation from Monocular Images: A Comprehensive Survey

Human Pose Estimation from Monocular Images: A Comprehensive Survey

Learning Gestures for Customizable Human-Computer Interaction in the Operating Room

Manifold Learning for ToF-based Human Body Tracking and Activity Recognition

Contact Info

Product

Resources

About