AI로 보는 저비용 체지방률 추정 이미지와 인체계측 데이터 활용

Multimodal AI for Body Fat Estimation: Computer Vision and Anthropometry with DEXA Benchmarks Rayan ALDajani Dept. of Electrical Engineering and Computer Science York University Toronto, Canada rayandajani21@gmail.com Abstract—Tracking body fat percentage is essential for ef- fective weight loss and health management, yet gold-standard methods such as DEXA scans [1], [2] are too expensive and rarely accessible for most people. This study aims to evaluate the feasibility of artificial intelligence (AI) models as low-cost alternatives using frontal body images and basic anthropometric data. The dataset consists of 535 samples: 253 cases with recorded anthropometric measurements (weight, height, neck, ankle, and wrist) and 282 images obtained via web scraping from Reddit posts self-reported body fat percentage values, some of which were stated to be derived from official DEXA scans. As no pub- licly available datasets exist for computer vision based body fat estimation, this dataset was compiled specifically for this study. Two approaches were developed: (1) ResNet-based image models, (2) regression models using measurements only. A multimodal fusion approach was proposed but could not be implemented due to the lack of paired datasets, and is identified as future work. The image-based model achieved a Root Mean Square Error (RMSE) of 4.44% and a Coefficient of Determination (R2) of 0.807. These results show that AI-assisted models and tools can give low-cost and accessible body fat estimates. This supports a future of consumer based weight loss and fitness apps. Index Terms—Body Fat Estimation, Computer Vision, Deep Learning, Multimodal Learning, Artificial Intelligence in Health- care. I. INTRODUCTION A. The Importance of Accessible Body Fat Estimation Body fat percentage (BF%) is an increasingly popular marker of health, fitness, and metabolic risk. Although ac- curate BF% tracking is important for measuring weight loss, athletic performance, and disease prevention, existing vali- dated methods are largely unavailable to the general public or too unreliable for weekly tracking. To enable widespread use in health and fitness, estimation must be both accurate and affordable. B. Limitations of Current Methods Dual-Energy X-ray Absorptiometry (DEXA), the gold- standard assessment of body fat [1], [2], provides an accurate body composition assessment and detailed health profiles not obtainable with any other technique. DEXA scans are expensive, time-consuming, and impractical for routine fitness tracking. Low-cost assessments, like the U.S. Navy method [3], which relies on neck, waist, and height measurements, are available for the individual’s use at home but are not consistent or accurate enough for appropriate monitoring of fitness and health. C. Role of Artificial Intelligence Recent advances in artificial intelligence (AI) and computer vision offer the possibility of combining affordability with reliability. Deep learning has been applied in obesity detection, pose estimation, and medical imaging [4]. However, no public datasets exist for image-based body fat estimation, leaving a gap in research and practical applications. D. Contribution of This Study This study addresses this gap by providing a new dataset of images and anthropometric data and testing two approaches: (1) ResNet-based regression on the image data only, (2) regression on the measurements only. Our results show that the image-based model yields the best estimates, with respect to DEXA-referenced cases, suggesting the possibility for AI- driven, low-cost consumer applications for monitoring body fat. II. BACKGROUND A. Anthropometric Models Low fidelity calculations, such as the Body Mass Index (BMI) or the U.S. Navy approach, are commonplace in the literature because they are inexpensive and easy to use; how- ever, they are not very accurate because they do not account for various nuances of body composition. B. Gold-standard Imaging Techniques DEXA provides accurate fat, lean mass, and density mea- surements, but is expensive and not widely used [1], [2], [5]. C. Artificial Intelligence in Health Applications Recent ML and vision advances enable health prediction tasks like disease detection and body composition estimation. These advances offer a cost-effective and scalable approach to DEXA-type measurements. arXiv:2511.17576v1 [cs.CV] 15 Nov 2025 III. RESULTS A. Dataset Description Two datasets were used. The image dataset consisted of 282 frontal body images with self-reported percentages of body fat (some referencing DEXA scans) and the anthropometric dataset consisted of 253 male records (Fisher and Johnson) where weight, height, and body circumferences were doc- umented (body fat obtained from underwater weighing and Siri’s equation). Paired samples were not available, thus each modality was considered independently. For both datasets, an 80/20 train/test split was applied with randomized shuffling and no subject overlap. A validation

This content is AI-processed based on open access ArXiv data.