End-to-End Recovery of Human Shape and Pose
By Angjoo Kanazawa, Michael J. Black, David W. Jacobs, Jitendra Malik. 2017
2,097 citations recorded by the source corpus.
We describe Human Mesh Recovery (HMR), an end-to-end framework for reconstructing a full 3D mesh of a human body from a single RGB image. In contrast to most current methods that compute 2D or 3D joint locations, we produce a richer and more useful mesh representation that is parameterized by shape and 3D joint angles. The main objective is to minimize the reprojection loss of keypoints, which allows our model to be trained using in-the-wild images that only have ground truth 2D annotations. However, the reprojection loss alone is highly underconstrained. In this work we address this problem by introducing an adversary trained to tell whether human body shape and pose parameters are real or not using a large database of 3D human meshes. We show that HMR can be trained with and without using any paired 2D-to-3D supervision. We do not rely on intermediate 2D keypoint detections and infer 3D pose and shape parameters directly from image pixels. Our model runs in real-time given a bounding box containing the person. We demonstrate our approach on various images in-the-wild and out-perform previous optimization-based methods that output 3D meshes and show competitive results on tasks such as 3D joint location estimation and part segmentation.
Subfields: Computational Mechanics, Computer Vision and Pattern Recognition, 3D Vision & Pose
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
- Gerard Pons-Moll
- Michael J. Black
- Andreas Geiger
- Christian Theobalt
- Michael Niemeyer
- Jitendra Malik
- Angjoo Kanazawa
- Yebin Liu
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Jitendra Malik
- Dacheng Tao
- Ming-Hsuan Yang
- Xiaogang Wang
- Trevor Darrell
- Andrea Vedaldi
- Andrew Zisserman
In 3D Vision & Pose
- Michael J. Black
- Andreas Geiger
- Jitendra Malik
- Michael Niemeyer
- Gerard Pons-Moll
- Angjoo Kanazawa
- Christian Theobalt
- Hanbyul Joo
Citations in the corpus
- Occupancy Networks: Learning 3D Reconstruction in Function Space (2018, background): “Moreover, they are only able to generate meshes with simple topology [70], require a reference template from the same object class [33, 41, 57] or cannot guarantee closed surfaces [26].”
- Learning Category-Specific Mesh Reconstruction from Image Collections (2018, background): “While some learning based methods do pursue mesh prediction [14,35,18,24], they also rely on 3D supervision which is only available for restricted classes or in a synthetic setting.”
- Compositional Human Pose Regression (2017, methodology): “, 2018), fitting a 3D model via optimization (Zhou et al., 2016c; Bogo et al., 2016; Zhou et al., 2017b; Mehta et al., 2017; Kanazawa et al., 2017), searching over a database (Chen and Ramanan, 2016), or generating multiple hypotheses (Jahangiri and Yuille, 2017).”
- FrankMocap: Fast Monocular 3D Hand and Body Motion Capture by Regression and Integration (2020, methodology): “Following the body pose estimation approaches [20, 25], the architecture of our hand module MH is composed of an”
- 3D Human Pose Estimation in Video With Temporal Convolutions and Semi-Supervised Training (2018, background): “While this paper and the discussed related work focus on reconstructing accurate 3D poses, a parallel line of research aims at recovering full 3D shapes of people from images [1, 23].”
- HuMoR: 3D Human Motion Model for Robust Pose Estimation (2021, background): “Deep learning approaches have shown success in regressing 3D shape and pose from a single image [47, 41, 71, 30, 29, 103, 20].”
- Review of Artificial Intelligence Techniques in Imaging Data Acquisition, Segmentation, and Diagnosis for COVID-19 (2020, methodology): “Unlike other related studies [46], they employ a hierarchical kinematic reasoning for each kinematic chain of the patient to iteratively refine the estimation of each anatomical”
- MotionBERT: A Unified Perspective on Learning Human Motion Representations (2022, methodology): “Note that most previous works [23,38,42,44,50,54,68,107] use more datasets other than COCO [57] during training, such as LSP [37], MPI-INF-3DHP [74], etc ., while we do not.”
- Articulation-Aware Canonical Surface Mapping (2020, background): “movement of parts, and a plethora of fitting based [4, 15] or prediction based [14, 36, 43] methods have been proposed to tackle this.”
- Humans in 4D: Reconstructing and Tracking Humans with Transformers (2023, methodology): “To extend Human Mesh Recovery over time, most methods use the basic backbone of HMR [29] and propose designs for the temporal encoder that fuses the per-frame features.”
- Unsupervised Learning of Probably Symmetric Deformable 3D Objects From Images in the Wild (2019, methodology): “For objects such as human bodies and human faces, some methods [26, 17, 60, 14] have learn to reconstruct from raw images, but starting from the knowledge of a predefined shape model such as SMPL [34] or Basel [43].”
- Neural Body Fitting: Unifying Deep Learning and Model Based Human Pose and Shape Estimation (2018, methodology): “[22] also integrate a learned prior on the space of poses.”
- Reconstructing Hands in 3D with Transformers (2023, methodology): “FrankMo-cap [49] is a good representative of this line of works which adopts a simple design, similar to HMR [27].”
- Learning to Regress 3D Face Shape and Expression From an Image Without 3D Supervision (2019, methodology): “This could go beyond current methods, like HMR [17], to learn about body shape.”
- Sim2real transfer learning for 3D human pose estimation: motion to the rescue (2019, methodology): “We outperform HMR [34], which trains on real-world motion capture datasets as well as real-world 2D images as a regularizer (ensuring that 3D poses are consistent with 2D annotations).”
- D3D-HOI: Dynamic 3D Human-Object Interactions from Videos (2021, neutral): “Most works in this area focus on separately re-constructing the human mesh [1, 10, 11, 15, 16, 27] or the individual 3D object [4, 18, 22, 31].”