Accurate 3D Human Pose in The Wild Using IMUs and a Moving Camera

By Timo von Marcard, Roberto Henschel, Michael J. Black, Bodo Rosenhahn, Gerard Pons-Moll. 2018

How researchers see this paper

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

Citations in the corpus

Transformers in Vision: A Survey (2021, neutral): “METRO [69] arXiv’20 MPJPE PA-MPJPE MPVE 3DPW [150] 77.”
Learning 3D Human Dynamics From Video (2018, methodology): “While more practical solutions are being introduced [47], 3D supervision is not available for millions of videos that are being uploaded daily on the Internet.”
End-to-End Human Pose and Mesh Reconstruction with Transformers (2020, background): “3DPW [49] is an outdoor-image dataset with 2D and 3D annotations.”
VIBE: Video Inference for Human Body Pose and Shape Estimation (2019, methodology): “We evaluate on 3DPW [66], MPI-INF3DHP [44], and Human3.”
Monocular human pose estimation: A survey of deep learning-based methods (2020, background): “Datasets play an important role in deep learning-baed human pose estimation.”
Tex2Shape: Detailed Full Human Body Geometry From a Single Image (2019, methodology): “We compare against ground truth on our own dataset and show qualitative results on 3DPW [51], DeepFashion [30, 31], and PeopleSnapshot [4] datasets.”
History Repeats Itself: Human Motion Prediction via Motion Attention (2020, methodology): “st set as each sequence consists of one actor performing one type of action. We then split the remaining parts of AMASS into training and validation data. 3DPW. The 3D Pose in the Wild dataset (3DPW) [21] consists of challenging indoor and outdoor actions. We only evaluate our model trained on AMASS”
On Self-Contact and Human Pose (2021, methodology): “We evaluate TUCH on the following three datasets: 3DPW [43], MPI-INF-3DHP [30], and 3DCP Scan.”
Deep Learning-based Human Pose Estimation: A Survey (2020, methodology): “Microsoft Common Objects in Context (COCO) Dataset [132] is the most widely used large-scale dataset.”
TAPIR: Tracking Any Point with per-frame Initialization and temporal Refinement (2023, methodology): “viewed from the front or back); thus most methods rely on large supervised datasets [2, 9, 44, 50, 67, 72], although some works track surface meshes [17, 62].”
Learning Trajectory Dependencies for Human Motion Prediction (2019, background): “6M) [10], the CMU mocap dataset1, and the 3DPW dataset [23].”
Mesh Graphormer (2021, methodology): “To get a fair comparison with earlier state-of-the-art [18, 22], we also use 3DPW [36] training data for 3DPW experiments.”
Learning to Reconstruct 3D Human Pose and Shape via Model-Fitting in the Loop (2019, unknown): “6M [11], MPI-INF-3DHP [22], LSP [12], and 3DPW [39].”
I2L-MeshNet: Image-to-Lixel Prediction Network for Accurate 3D Human Pose and Mesh Estimation from a Single RGB Image (2020, background): “3DPW [29] contains 60 video sequences captured mostly in outdoor conditions.”
Human Motion Diffusion as a Generative Prior (2023, methodology): “We train and evaluate ComMDM with the 3DPW dataset [Von Marcard et al. 2018], which contains 27 two-person motion sequences annotated with SMPL joints.”
MotionBERT: A Unified Perspective on Learning Human Motion Representations (2022, unknown): “6M [38] and 3DPW [113] datasets and additionally add COCO [61] dataset during training following [46, 58, 114].”
Pose2Mesh: Graph Convolutional Network for 3D Human Pose and Mesh Recovery from a 2D Human Pose (2020, methodology): “fter further alignment (i.e., Procrustes analysis (PA) [16]). JM is used for the estimated joint coordinates. We only evaluate 14 joints out of 17 estimated joints following [26,30,31,50]. 3DPW. 3DPW [38] is captured from in-the-wild and contains 3D body pose and mesh annotations. It consists of 51K”
FrankMocap: Fast Monocular 3D Hand and Body Motion Capture by Regression and Integration (2020, background): “6M [14] and in-the-wild datasets such as COCO [27] or 3DPW [54].”
Sim2real transfer learning for 3D human pose estimation: motion to the rescue (2019, unknown): “3D human pose estimation is a rare exception to this trend: manual annotation is almost impossible, yet a small dataset of in-the-wild data is available for evaluation due to clever use of external sensors [87], called 3D Poses in the Wild (3DPW).”