Pose-independent 3D Anthropometry from Sparse Data

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.

• The paper presents a novel technique to accurately measure human body dimensions using sparse 3D landmark data from any pose, unlike traditional methods requiring static postures and dense scans.
• The proposed method identifies pose-independent features from landmark distances and uses a Multilayer Perceptron (MLP) to predict body measurements, offering flexibility and resilience to noise.
• This research demonstrates performance comparable to dense data methods, enabling more flexible and accessible 3D anthropometry with practical implications for industries like medicine, fashion, and ergonomics.

Pose-independent 3D Anthropometry from Sparse Data

The paper "Pose-independent 3D Anthropometry from Sparse Data" presents a novel technique for obtaining precise human body measurements from sparse 3D landmark data, rendering the traditional reliance on dense geometrical data and static postures unnecessary. This research is significant in the field of digital anthropometry, with implications for industries such as medicine, fashion, ergonomics, and more.

Summary

The core concept of this work is to decouple the anthropometric measurement process from the limitations imposed by the static A-pose, which is traditionally required for accurate 3D body scanning. Instead of using dense 3D data, which necessitates expensive equipment and heavily restricts subject mobility, this paper advocates for a model that predicts body measurements accurately using only sparse data from any pose. This innovation allows for measurements even in circumstances where maintaining a fixed pose is difficult, such as for individuals with physical disabilities.

The authors identify pose-independent features by analyzing landmark distances from a dataset of scans in various poses. These invariant features constitute the input for their predictive model, which is based on a Multilayer Perceptron (MLP). This approach contrasts with existing methods that often require dense data or meticulously fitted templates that do not generalize well to new poses. The resultant method reportedly maintains performance quality comparable to existing technologies that necessitate dense scans.

Results

The paper details a robust series of comparisons with both dense data-driven methods and existing sparse methods. Notably, the proposed method not only matches but often surpasses these in terms of efficiency and flexibility. Specifically, it retains high performance across different poses and is resilient to typical landmarking noise—posing a significant advancement over contemporaries such as those relying on the CAESAR dataset and other established benchmarks.

By leveraging a large-scale posed dataset, the authors empirically establish that their method maintains predictability well within acceptable error margins as defined by manual anthropometric measurement standards. Furthermore, they illuminate the ambiguity inherent in sparse landmark data, where significant shape variations can exist for near-identical landmark positions, thereby underscoring the complexity their model aims to transcend.

Implications and Future Work

This research paves the way for more flexible and accessible 3D body measurement technologies, particularly beneficial for scenarios where traditional scanning setups are impractical. Practically, industries reliant on anthropometry can benefit from reduced costs and expanded accessibility, as lower-cost scanning (e.g., mobile devices) becomes feasible.

Theoretically, the method challenges and extends current paradigms, encouraging further explorations into minimalistic data dependencies in 3D reconstruction and measurement. Future research could focus on refining these models in diverse populations and integrating more nuanced data features (e.g., texture, biometric markers) to enhance precision further.

This work also sets a precedent for open-source collaboration by sharing its methodology and evaluation protocol, inviting the research community to engage with and build upon the framework presented.

In conclusion, by demonstrating accurate body measurement from sparse, pose-independent data, this paper significantly contributes to the evolution of 3D anthropometry, making it more inclusive and adaptable across different applications and environments.