Real time motion capture using a single time-of-flight camera

By Varun Ganapathi, Christian Plagemann, Daphne Koller, Sebastian Thrun. 2010

425 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Control and Systems Engineering

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

In Control and Systems Engineering

Citations in the corpus

Space-Time Representation of People Based on 3D Skeletal Data: A Review (2016, methodology): “Many other MoCap datasets were also released, including the Pictorial Human Spaces [102], CMU Multi-Modal Activity (CMU-MMAC) [103] Berkeley MHAD [104], Standford ToFMCD [26], HumanEva-I [105], and HDM05 MoCap [106] datasets.”
Accurate 3D pose estimation from a single depth image (2011, methodology): “In order to compare with this ground truth, we choose a set of patches on our mesh model to approximate the position of markers according to their markers setup, as was done in [9].”
Human detection using depth information by Kinect (2011, background): “Many algorithms have been proposed to address the problem of pose estimation and motion capture from range images [5] [7] [11] [16].”
Functional categorization of objects using real-time markerless motion capture (2011, background): “[12] propose using body part detectors [23] for full body motion capture.”
DART: Dense Articulated Real-Time Tracking (2014, methodology): “…tracking, e.g. by Ballan et al. [1], who used fingernail detectors to help constrain a gradient descent optimization, by Ganapathi et al. [11], who used a gradient-based optimizer in conjunction with body part recognition for human body tracking, and by Qian et al. [26], who combine PSO, a…”
The Vitruvian manifold: Inferring dense correspondences for one-shot human pose estimation (2012, methodology): “(b) Comparison on the Stanford dataset [15].”
Human3.6M: Large Scale Datasets and Predictive Methods for 3D Human Sensing in Natural Environments (2014, background): “of time-of-flight, infrared or structured light sensors [39], [40].”
Free-Viewpoint Video of Human Actors Using Multiple Handheld Kinects (2013, background): “Monocular depth-based body tracking with improved accuracy at near-real-time is feasible by combination of pose optimization with body part detection [41].”
Articulated and Generalized Gaussian Kernel Correlation for Human Pose Estimation (2016, methodology): “The hybrid methods [8]–[10], [19] take advantage of the complementary nature of the two kinds of approaches that involve querying or training data and useful data-driven detectors to assist the model-based optimization process.”
Metric Regression Forests for Correspondence Estimation (2015, background): “…the69 possible variation in the training data, but yields a set of inde-70 pendent local pose cues that are unlikely to respect kinematic71 constraints.72 To overcome this, recent work has fit a generative model73 to these cues (Ganapathi et al. 2010; Baak et al. 2011; Taylor74 et al. 2012).”
Efficient Human Pose Estimation from Single Depth Images (2013, background): “Traditional human body tracking algorithms [8], [9], [1], [10], [6], [11] infer a complete skeleton by exploiting kinematic constraints and can achieve high frame-rates by using temporal coherence from frame-to-frame.”
A data-driven approach for real-time full body pose reconstruction from a depth camera (2011, methodology): “Only for sequence 24 we perform worse than [10].”
Home 3D body scans from noisy image and range data (2011, methodology): “Existing commercial and research systems solve the problem of rough body pose estimation from this type of sensor data [12, 23] but, to date, there are no methods for accurate body shape estimation.”
Towards Viewpoint Invariant 3D Human Pose Estimation (2016, methodology): “The second challenge is that existing real-world depth datasets are often small in size, both in terms of number of frames and number of classes [21,20].”
Capturing the human figure through a wall (2015, background): “…– including Kinect, multi-view cameras, moving cameras, and time-of-ﬂight cameras – and require a direct line-of-sight from the sensor to the person’s body, e.g., [Shotton et al. 2013; Ganapathi et al. 2010; Hasler et al. 2009b; Gall et al. 2009; Vlasic et al. 2008; Poppe 2007; Ye et al. 2014].”
A combined pose, object, and feature model for action understanding (2012, background): “[9, 30]) have been able to perform accurate tracking of human pose in real-time based on the depth readings alone.”