Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Pose-independent 3D Anthropometry from Sparse Data

David Bojanić, Stefanie Wuhrer, Tomislav Petković, Tomislav Pribanić

TL;DR

This paper tackles the problem of estimating 3D body measurements without requiring the subject to remain in the fixed A-pose, enabling pose-invariant anthropometry from sparse landmarks. It introduces a learning approach that uses 70 landmarks plus 158 pose-independent features to predict 11 standard measurements in the A-pose, trained on CAESAR and evaluated on CAESAR, DYNA, and 4DHumanOutfit data. The method achieves accuracy comparable to dense-geometry approaches while enabling estimation from arbitrary poses, showing robustness to landmark noise and clothing, and it is released as open-source to foster reproducibility and broader use. This work significantly lowers the data-collection burden and expands applicability to subjects who cannot maintain an A-pose, with practical impact for health, fashion, and ergonomics.

Abstract

3D digital anthropometry is the study of estimating human body measurements from 3D scans. Precise body measurements are important health indicators in the medical industry, and guiding factors in the fashion, ergonomic and entertainment industries. The measuring protocol consists of scanning the whole subject in the static A-pose, which is maintained without breathing or movement during the scanning process. However, the A-pose is not easy to maintain during the whole scanning process, which can last even up to a couple of minutes. This constraint affects the final quality of the scan, which in turn affects the accuracy of the estimated body measurements obtained from methods that rely on dense geometric data. Additionally, this constraint makes it impossible to develop a digital anthropometry method for subjects unable to assume the A-pose, such as those with injuries or disabilities. We propose a method that can obtain body measurements from sparse landmarks acquired in any pose. We make use of the sparse landmarks of the posed subject to create pose-independent features, and train a network to predict the body measurements as taken from the standard A-pose. We show that our method achieves comparable results to competing methods that use dense geometry in the standard A-pose, but has the capability of estimating the body measurements from any pose using sparse landmarks only. Finally, we address the lack of open-source 3D anthropometry methods by making our method available to the research community at https://github.com/DavidBoja/pose-independent-anthropometry.

Pose-independent 3D Anthropometry from Sparse Data

TL;DR

This paper tackles the problem of estimating 3D body measurements without requiring the subject to remain in the fixed A-pose, enabling pose-invariant anthropometry from sparse landmarks. It introduces a learning approach that uses 70 landmarks plus 158 pose-independent features to predict 11 standard measurements in the A-pose, trained on CAESAR and evaluated on CAESAR, DYNA, and 4DHumanOutfit data. The method achieves accuracy comparable to dense-geometry approaches while enabling estimation from arbitrary poses, showing robustness to landmark noise and clothing, and it is released as open-source to foster reproducibility and broader use. This work significantly lowers the data-collection burden and expands applicability to subjects who cannot maintain an A-pose, with practical impact for health, fashion, and ergonomics.

Abstract

3D digital anthropometry is the study of estimating human body measurements from 3D scans. Precise body measurements are important health indicators in the medical industry, and guiding factors in the fashion, ergonomic and entertainment industries. The measuring protocol consists of scanning the whole subject in the static A-pose, which is maintained without breathing or movement during the scanning process. However, the A-pose is not easy to maintain during the whole scanning process, which can last even up to a couple of minutes. This constraint affects the final quality of the scan, which in turn affects the accuracy of the estimated body measurements obtained from methods that rely on dense geometric data. Additionally, this constraint makes it impossible to develop a digital anthropometry method for subjects unable to assume the A-pose, such as those with injuries or disabilities. We propose a method that can obtain body measurements from sparse landmarks acquired in any pose. We make use of the sparse landmarks of the posed subject to create pose-independent features, and train a network to predict the body measurements as taken from the standard A-pose. We show that our method achieves comparable results to competing methods that use dense geometry in the standard A-pose, but has the capability of estimating the body measurements from any pose using sparse landmarks only. Finally, we address the lack of open-source 3D anthropometry methods by making our method available to the research community at https://github.com/DavidBoja/pose-independent-anthropometry.
Paper Structure (22 sections, 2 equations, 4 figures, 6 tables)

This paper contains 22 sections, 2 equations, 4 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. 3D anthropometry
  4. 3D anthropometry from posed subjects
  5. Method
  6. Datasets
  7. CAESAR
  8. Posed training set.
  9. Standing test set.
  10. Noisy test set.
  11. Sitting test set.
  12. Posed test set.
  13. Dynamic test set
  14. Clothed test set
  15. Experiments
  16. ...and 7 more sections

Figures (4)

  • Figure 1: We propose a new approach to estimate body measurements from 3D landmark locations on a body in any given pose. We identify pose-independent features that have an impact on the measurements by analyzing the landmarks of a large database of posed scans. The $11$ body measurements listed on the right, are estimated from the $70$ landmarks listed on the left, along with $158$ pose-independent features in the middle.
  • Figure 2: We show the landmark distances of the same subject in two different poses. As can be seen, the same landmark distances $d_1$ and $d'_1$ do not significantly change between the two poses, observing an absolute difference of only $0.78$ cm. On the contrary, the landmark distances $d_2$ and $d'_2$ change significantly because of the articulated deformation of the leg. In this case, the absolute difference between $d_2$ and $d'_2$ is $36.75$ cm. In our method we only use landmark distances where the median difference from its appropriate A-pose distance is less than $1$ cm, such as the distance $d_1$.
  • Figure 3: Some of the poses used to create the posed test set. We augment the CAESAR dataset by reposing the subjects in various poses obtained from the AIST aist_dataset dataset.
  • Figure 4: We answer the question whether different body shapes, with different measurements, can share the same landmarks. The relationship between the landmarks and measurements is linear since it is modeled with SMPL, and shows an approximation of the relationship that would be obtained with real data. As can be seen, the chest and hip circumference measurements change faster than the landmark distances, indicating that two subjects can have close-by landmark coordinates with different chest and hip circumferences.