← ArchivePaper2022
SUPR: A Sparse Unified Part-Based Human Representation
Ahmed A. A. Osman, Timo Bolkart, Dimitrios Tzionas, Michael J. Black
SUPR jointly trains a full-body and part-specific models from 1.2 million scans, introducing a novel kinematic foot model with contact-aware deformations.
Abstract
Statistical 3D shape models of the head, hands, and fullbody are widely used in computer vision and graphics. Despite their wide use, we show that existing models of the head and hands fail to capture the full range of motion for these parts. Moreover, existing work largely ignores the feet, which are crucial for modeling human movement and have applications in biomechanics, animation, and the footwear industry. The problem is that previous body part models are trained using 3D scans that are isolated to the individual parts. Such data does not capture the full range of motion for such parts, e.g. the motion of head relative to the neck. Our observation is that full-body scans provide important information about the motion of the body parts. Consequently, we propose a new learning scheme that jointly trains a full-body model and specific part models using a federated dataset of full-body and body-part scans. Specifically, we train an expressive human body model called SUPR (Sparse Unified Part-Based Human Representation), where each joint strictly influences a sparse set of model vertices. The factorized representation enables separating SUPR into an entire suite of body part models. Note that the feet have received little attention and existing 3D body models have highly under-actuated feet.
How to read this
- Category
- Method / model: a unified part-based human body model
- Contributions
- SUPR, an expressive body model where each joint strictly influences a sparse set of vertices
- A learning scheme that jointly trains a full-body model and separable part models from a federated dataset of full-body and body-part scans
- A novel kinematic foot model with contact-aware deformations, and the ability to separate SUPR into a suite of body-part models
- Context
- Builds directly on sparse articulated body modeling (Osman et al., STAR, 2020) and addresses the limitation that part-only scans miss the full range of motion that full-body scans reveal.Builds on: STAR: Sparse Trained Articulated Human Body Regressor
- Correctness
- The key claim is that jointly training on full-body plus part scans captures motion (e.g. head relative to neck, foot contact) that isolated-part training misses; it is trained on a large scan corpus, but expressiveness for unusual anatomies or poses outside the captured distribution remains a fair caution.
- Clarity
- Reasonably accessible for readers familiar with parametric body models; a first pass conveys the federated-training idea, a second pass for the factorization and foot model.
- How to read it
- First pass for the motivation and the federated joint-training scheme; second pass on the sparse joint-to-vertex factorization and the kinematic contact-aware foot model if you use or extend body models like SMPL or STAR.
Builds on
Built upon by
Nothing yet.
Related work
- STAR: Sparse Trained Articulated Human Body Regressor 2020 / Eurographics
- Expressive Body Capture: 3D Hands, Face, and Body from a Single Image 2019 / CVPR
- ATLAS: Decoupling Skeletal and Shape Parameters for Expressive Parametric Human Modeling 2025 / ICCV
- NIMBLE: A Non-rigid Hand Model with Bones and Muscles 2022 / TOG
Keywords
This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →