Using Linking Features in Learning Non-parametric Part Models

By Leonid Karlinsky, Shimon Ullman. 2012

53 citations recorded by the source corpus.

Subfields: 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 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.”
Convolutional Pose Machines (2016, background): “Non-tree models [8, 16, 19, 33, 42] incorporate interactions that introduce loops to augment the tree structure with additional edges that capture symmetry, occlusion and long-range relationships.”
Articulated Pose Estimation by a Graphical Model with Image Dependent Pairwise Relations (2014, methodology): “ generalizes better to the Buffy dataset. Method U.arms L.arms Mean Ours* 96.8 89.0 92.9 Ours* strict 94.5 84.1 89.3 Yang[27] 97.8 68.6 83.2 Yang[27] strict 94.3 57.5 75.9 Sapp[21] 95.3 63.0 79.2 FLPM[11] 93.2 60.6 76.9 Eichner[5] 93.2 60.3 76.8 Table3: Cross-dataset PCP results on Buffy test subset”
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.”
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 Deep Consensus Voting (2016, background): “By adding this mid-point constraint, and using it as a linking feature [11], we get a more reliable binary term which also looks at the appearance of the space between the two respective keypoints.”
Knowledge-Guided Deep Fractal Neural Networks for Human Pose Estimation (2017, methodology): “On the other hand, there are non-tree models [23], [24] to incorporate interactions that introduce loops to augment the tree structure with additional edges that capture symmetry, occlusion and long-range relationships.”
Human Pose Estimation via Convolutional Part Heatmap Regression (2016, background): “Recent work includes sophisticated extensions like mixture, hierarchical, multimodal and strong appearance models [2–6], non-tree models [7, 8] as well as cascaded/sequential prediction models like pose machines [9].”
Human Pose Estimation Using a Joint Pixel-wise and Part-wise Formulation (2013, background): “Recent work has attempted to overcome these problems, for example by enforcing consistency of ensembles of parts [15, 23, 29] or eschewing the pictorialstructure formulation by directly learning poselets for human parts tightly clustered in both appearance and configuration spaces [3].”