Total Capture: 3D Human Pose Estimation Fusing Video and Inertial Sensors

By Matthew Trumble, Andrew Gilbert, Charles Malleson, Adrian Hilton, John Philip Collomosse. British Machine Vision Conference, 2017

339 citations recorded by the source corpus.

Subfields: Biomedical Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/22cd49c5271b962575f145d5e3b5119496cc7f5f

How researchers see this paper

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

In Biomedical Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Single-Shot Multi-person 3D Pose Estimation from Monocular RGB (2017, background): “3D Pose Datasets : Existing pose datasets are either for a single person in 3D [18, 51, 63, 64, 33] or multi-person with only 2D pose annotations [3, 29].”
Deep 3D human pose estimation: A review (2021, positive): “Recently, some methods (Trumble et al., 2017; von Marcard et al., 2018) have been proposed to solve this problem, and to a certain extent, these methods have made some progress.”
Recovering Accurate 3D Human Pose in the Wild Using IMUs and a Moving Camera (2018, methodology): “Another recent dataset is TotalCapture [39], which features synchronized video, marker-based ground-truth poses and IMUs.”
MotionBERT: A Unified Perspective on Learning Human Motion Representations (2022, methodology): “AMASS [71] integrates most existing marker-based Mocap datasets [1,2,4,10,14,17,32,35,48,65,67,72,79,90,103,104] and parameterizes them with a common representation.”
Monocular human pose estimation: A survey of deep learning-based methods (2020, methodology): “Datasets play an important role in deep learning-baed human pose estimation.”
Synthetic Data for Deep Learning (2019, background): “• basic human-related tasks are very important in practice, so there already exist large real datasets for face recognition [84, 220, 355], pose estimation [24, 274, 373, 600], and other problems, which often limits synthetic”
CLIFF: Carrying Location Information in Full Frames into Human Pose and Shape Estimation (2022, neutral): “However, it costs a lot of time and effort to obtain these 3D annotations, using multi-view motion capture systems [14,33] or a set of IMU devices [50,32].”
Deep Learning-based Human Pose Estimation: A Survey (2020, background): “3M frames Yes No Yes Yes TotalCapture [235] 2017 Marker-based MoCap with IMUs Indoor 5 subjects, 5 actions, 1.”
MotionDiffuse: Text-Driven Human Motion Generation With Diffusion Model (2022, background): “, 2018), 3D keypoints (Ionescu et al., 2013; Joo et al., 2015; Mehta et al., 2017; Trumble et al., 2017; Li et al., 2021) and statistical model parameters (Yu et al.”
AMASS: Archive of Motion Capture As Surface Shapes (2019, background): “There are many motion capture datasets [3, 9, 10, 21, 25, 31, 30, 39, 34, 42, 43, 45], as well as several attempts to aggregate such datasets into larger collections [20, 27, 29].”
HuMoR: 3D Human Motion Model for Robust Pose Estimation (2021, methodology): “For training the CVAE, we use the recommended training split (save for TCD Hands [14] which contains mostly hand motions): CMU [6], MPI Limits [2], TotalCapture [35], Eyes Japan [20], KIT [24], BMLrub [34], BMLmovi [10], EKUT [24], and ACCAD [1].”