Dynamic FAUST: Registering Human Bodies in Motion

By Federica Bogo, Javier Romero, Gerard Pons-Moll, Michael J. Black. Computer Vision and Pattern Recognition, 2017

432 citations recorded by the source corpus.

Subfields: Computational Mechanics, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/9b84ae8241d4fe5a41d996af025d8a6e36d1666a

How researchers see this paper

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

In Computational Mechanics

In Computer Vision and Pattern Recognition

Citations in the corpus

Video Based Reconstruction of 3D People Models (2018, methodology): “For quantitative evaluation, we used two publicly available datasets consisting of 3D scan sequences of humans in motion: with minimal clothing (MC) (DynamicFAUST [8]) and with clothing (BUFF [80]).”
Shape As Points: A Differentiable Poisson Solver (2021, methodology): “We consider three different types of 3D inputs: point clouds sampled from synthetic meshes [74] with Gaussian noise, real-world scans [68], and high-resolution raw scans of humans with comparably little noise [6].”
Implicit Geometric Regularization for Learning Shapes (2020, methodology): “, triangle soups) of a human, S , from the D-Faust (Bogo et al., 2017) dataset.”
AMASS: Archive of Motion Capture As Surface Shapes (2019, methodology): “As an additional evaluation, we performed another experiment using one of the largest publicly available 4Dscan datasets, Dynamic FAUST (DFAUST) [2].”
ClothCap (2017, background): “Moreover, adding appearance information to the tracking framework as in (Bogo et al. 2017) should result in even more stable tracking results.”
Multi-chart generative surface modeling (2018, methodology): “For human body shapes we used datasets of human bodies [7, 38] consisting of different humans in a collection of poses as input to our 3D GAN framework.”
Occupancy Flow: 4D Reconstruction by Learning Particle Dynamics (2019, methodology): “We use the Dynamic FAUST (D-FAUST) [9] dataset which contains scans and meshes for 129 sequences of 10 real humans performing various motions such as “punching”, “chicken wings”, or “jumping jacks”.”
Monocular human pose estimation: A survey of deep learning-based methods (2020, methodology): “3D human body shape scans, such as SCAPE (Anguelov et al., 2005), INRIA4D (INRIA4D, accessed on 2019) and FAUST (Bogo et al., 2014, 2017), 3D human body surface cloud points with time of ﬂight (TOF) depth sensors (Shahroudy et al., 2016), 3D human body reﬂective markers capture with motion capture s”
On Learning Sets of Symmetric Elements (2020, methodology): “To generate the data, we cropped 7-frame-long sequences from the Dynamic Faust dataset (Bogo et al., 2017) in which the shapes are given as triangular meshes.”
ABC: A Big CAD Model Dataset for Geometric Deep Learning (2018, methodology): “Further datasets are available for large scenes [51, 23], mesh registration [14] and 2D/3D alignment [12].”
Learning Representations and Generative Models for 3D Point Clouds (2017, methodology): “In addition to ShapeNet core which contains man-made only objects, we have experimented with the D-FAUST dataset of (Bogo et al., 2017) that contains meshes of human subjects.”
Self-supervised Learning of Dense Shape Correspondence (2018, background): “For example, despite being restricted to a single shape class (human bodies), the MPI FAUST scanning and labeling system [8] required substantial manual labor and considerable financial costs.”
GRAB: A Dataset of Whole-Body Human Grasping of Objects (2020, methodology): “To evaluate GrabNet generated grasps, we perform a user study through AMT [1].”
Monocular Total Capture: Posing Face, Body, and Hands in the Wild (2018, methodology): “Some work [16, 10] also uses optical flow to align rendered 3D human models.”
NASA: Neural Articulated Shape Approximation (2019, methodology): “The meshes are obtained from the “DFaust” [5] and “Transitions” sub-datasets of AMASS [35], as detailed in Section 6.”
Deformable Shape Completion with Graph Convolutional Autoencoders (2017, methodology): “The VAE is trained on registered 4D scans from the DFAUST dataset [6] comprising 10 human subjects performing 14 different activities.”
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.”
SAL: Sign Agnostic Learning of Shapes From Raw Data (2019, methodology): “In the main experiment of this paper we trained on the D-Faust scan dataset [7], consisting of approximately 41k raw scans of 10 humans in multiple poses(2).”