Measure Locally, Reason Globally: Occlusion-sensitive Articulated Pose Estimation

By Leonid Sigal, Michael J. Black. 2006

259 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

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

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Realtime Multi-person 2D Pose Estimation Using Part Affinity Fields (2016, background): “…models [10, 30, 4, 3, 26, 40, 17], which parametrically encode the spatial relationship between adjacent parts following a kinematic chain, or non-tree models [37, 31, 20, 18, 8] that augment the tree structure with additional edges to capture occlusion, symmetry, and long-range relationships.”
Articulated Human Detection with Flexible Mixtures of Parts (2013, background): “However, such methods may overfit to the statistics of a particular dataset, as warned by [18], [23].”
Pose Machines: Articulated Pose Estimation via Inference Machines (2014, methodology): “Pictorial structure models with non-tree interactions have been employed [12,13,14,15] to estimate pose in a single image.”
An Efficient Convolutional Network for Human Pose Estimation (2016, methodology): “Complex Non-tree Models [23] have also been used that model spatial dependencies between non-neighbouring parts and then rely on loopy belief propagation with approximate to reason about different body joint positions in the image to make inference tractable.”
Progressive search space reduction for human pose estimation (2008, background): “4) is most closely related that of [23, 24], but our framework is fully automatic (it does not need any manual initialization or background subtraction).”
OpenPose: Realtime Multi-Person 2D Pose Estimation Using Part Affinity Fields (2018, methodology): “The spatial model for articulated pose is either based on tree-structured graphical models [7], [8], [9], [10], [11], [12], [13], which parametrically encode the spatial relationship between adjacent parts following a kinematic chain, or non-tree models [14], [15], [16], [17], [18] that augment the ”
Human Pose Estimation Using Body Parts Dependent Joint Regressors (2013, background): “To prevent this, additional constraints between the limbs [31, 25, 17, 34] or even a fully connected graphical model [2, 36] have been proposed.”
Body Parts Dependent Joint Regressors for Human Pose Estimation in Still Images (2014, background): “Index Terms—Human pose estimation, fashion, random forest, regression, classification F”
We Are Family: Joint Pose Estimation of Multiple Persons (2010, background): “the upper arms must be attached to the torso) and, in a few works, also other relations such as self-occlusion constraints [7] or the coordination between parts [17] (e.”
HumanEva: Synchronized Video and Motion Capture Dataset and Baseline Algorithm for Evaluation of Articulated Human Motion (2010, methodology): “2003) have been developed for tracking people from single (Agarwal and Triggs 2004a, 2004b; Felzenszwalb and Huttenlocher 2005; Hua et al. 2005; Lan and Huttenlocher 2005; Mori 2005, Mori et al. 2004; Ramanan et al. 2005; Ramanan and Forsyth 2003; Ren et al. 2005; Ronfard et al. 2002; Sigal and Blac”
Pictorial structures revisited: People detection and articulated pose estimation (2009, methodology): “[20] also uses non-tree models to improve occlusion handling, but still relies on simple features, such as color.”
Vision-based human motion analysis: An overview (2007, background): “Sigal and Black [97] introduced occlu-sion-sensitive image likelihoods, which introduces loops in the graphical model.”
HumanEva: Synchronized Video and Motion Capture Dataset for Evaluation of Articulated Human Motion (2006, methodology): “As a convention from previous methods [3, 15, 37, 36] that have already used this error metric we compute the 3D error in millimeters ( mm ) and 2D error directly in the image in pixels ( pix ).”
Max Margin AND/OR Graph learning for parsing the human body (2008, background): “Sigal and Black [17] address the occlusion problem by enhancing the ability of appearance modeling.”
People-tracking-by-detection and people-detection-by-tracking (2008, background): “Sigal and Black [21], for example, integrate occlusion reasoning into a 2D articulated tracking model.”
Putting the pieces together: Connected Poselets for human pose estimation (2011, background): “Loopy belief propagation models [25, 29, 27] and fully connected models [28] require approximations to infer the marginals.”