Related Experiment Video
Updated: Jun 16, 2026
14:25
Determining 3D Flow Fields via Multi-camera Light Field Imaging
Published on: March 6, 2013
16.7K
Modification and refinement of three-dimensional reconstruction to estimate body volume from a simulated
Chuang-Yuan Chiu1,2, Marcus Dunn1,2, Ben Heller1,2
1Sports Engineering Research Group Sheffield Hallam University Sheffield UK.
Obesity Science & Practice
|April 10, 2023
Summary
This study adapted a machine learning technique to estimate body volumes (BV) from images, achieving high accuracy for individuals with a body mass index (BMI) below 30. The method offers a convenient tool for obesity assessments in various settings.
Area of Science:
- Biomedical Engineering
- Computer Science
- Anthropometry
Background:
- Body volumes (BV) are crucial for body composition and obesity assessment.
- Traditional BV estimation methods are often impractical.
- Machine learning offers potential for non-invasive body measurement from images.
Purpose of the Study:
- To adapt and evaluate the STRAPS machine learning technique for estimating body volumes (BV).
- To assess the accuracy of BV estimation using a novel image-based approach.
Main Methods:
- Applied the STRAPS technique to generate 3D models from 2D images.
- Scaled 3D models with body stature and used regression for BV estimation.
- Validated against a large 3D scan dataset (n=4318).
Main Results:
- Achieved relative standard errors of estimation below 7% for BV.
- Demonstrated higher accuracy for individuals with BMI < 30 kg/m² (1.8-1.9%) compared to BMI ≥ 30 kg/m² (2.4-6.9%).
Conclusions:
- The developed method is suitable for BV estimation in males and females with BMI < 30 kg/m².
- This technique can facilitate at-home or clinical obesity assessments.
