The Vitruvian manifold: Inferring dense correspondences for one-shot human pose estimation

By Jonathan Taylor, Jamie Shotton, Toby Sharp, Andrew William Fitzgibbon. 2012

320 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

DensePose: Dense Human Pose Estimation in the Wild (2018, background): “The task of establishing dense correspondences from an image to a surface-based model has been addressed mostly in the setting where a depth sensor is available [43, 34, 46].”
DISCO Nets : DISsimilarity COefficients Networks (2016, background): “[26], that are, the Mean Joint Euclidean Error (MeJEE), the Max Joint Euclidean Error (MaJEE) and the Fraction of Frames within distance (FF).”
Space-Time Representation of People Based on 3D Skeletal Data: A Review (2016, methodology): “Another widely investigated methodology to construct 3D human skeleton models from depth imagery is based on nearest-neighbor matching [52], [57], [58], [59], [60], [63].”
Dense Human Body Correspondences Using Convolutional Networks (2015, background): “Regression forests [41, 31] can also be used to infer geometric locations from depth image, however such methods has not yet achieve comparable accuracies with state-of-the-art registration method on full or partial data [10].”
Flowing ConvNets for Human Pose Estimation in Videos (2015, neutral): “More recent work has tackled pose estimation holistically: initially with Random Forests on depth data [12, 29, 31, 34] and RGB [3, 4, 24], and most recently with convolutional neural networks.”
Deeper Depth Prediction with Fully Convolutional Residual Networks (2016, background): “Moreover, the availability of reasonably accurate depth information is well-known to improve many computer vision tasks with respect to the RGB-only counterpart, for example in reconstruction [23], recognition [26], semantic segmentation [5] or human pose estimation [35].”
A-NeRF: Articulated Neural Radiance Fields for Learning Human Shape, Appearance, and Pose (2021, background): “NeRF could then learn without change in the 3D space of the rest pose as done before for surfaces [64].”
DART: Dense Articulated Real-Time Tracking (2014, methodology): “Discriminative approaches have proven extremely successful for the problem of human pose estimation [29, 31], and similar approaches have also been applied to hand tracking [17].”
Dense Object Nets: Learning Dense Visual Object Descriptors By and For Robotic Manipulation (2018, background): “Most work on learning dense correspondences [9, 10, 11] requires absolute labels which must be annotated manually.”
Scene Coordinate Regression Forests for Camera Relocalization in RGB-D Images (2013, methodology): “[32] use a regression forest to infer correspondences between each depth image pixels and points in a canonical articulated 3D human mesh.”
Bayesian Image Based 3D Pose Estimation (2016, methodology): “Breaking down the skeleton into groups is motivated by the idiosyncratic motion of body parts, and to appraise this fact we use the Da Vinci’s Vitruvian pose as the reference skeleton conﬁguration, adapting an idea of [30].”
From Point Clouds to Mesh Using Regression (2017, background): “However, due to their large computational requirements to train large 3D models [21] and the slow evaluation, we decided to use regression forests [3], which were already successfully applied to similar contextbased learning problems [28, 15].”
Human Action Recognition From Various Data Modalities: A Review (2020, unknown): “[167], [168] and depth data [169], [170], [171], [172], [173], [174].”
User-Specific Hand Modeling from Monocular Depth Sequences (2014, methodology): “To achieve this, we adapt the ‘Vitruvian Manifold’ method of [27] to directly predict correspondences from the input image.”
Articulated and Generalized Gaussian Kernel Correlation for Human Pose Estimation (2016, result): “We evaluate our articulated pose tracking algorithm on two depth benchmark datasets, i.e., [8] (body) and [19] (hand), which shows that the accuracy of pose estimation is competitive compared to the best results reported so far [6], [10], [20].”
Hand Pose Estimation through Weakly-Supervised Learning of a Rich Intermediate Representation (2015, background): “In [29], a random forest is learned, which performs different tasks at different levels: viewpoint clustering in higher lvels, part segmentation in middle levels, and joint regression in lower levels.”
Latent Regression Forest: Structured Estimation of 3D Articulated Hand Posture (2014, methodology): “Additionally, in Figures 8 (c) and (f), we compare all methods using the more challenging metric proposed in [18].”
Metric Regression Forests for Correspondence Estimation (2015, methodology): “As 208 we will show, MSIG outperforms the PARTS (Shotton et al. 209 2011; Taylor et al. 2012) and standard regression (Girshick 210 et al. 2011) objectives, and can be computed efficiently in 211 linear time.”
Cascaded hand pose regression (2015, methodology): “Recent years have seen rapid progress and significant success of human body pose estimation [18, 29, 1, 19, 36] using consumer depth sensors.”