AUTOMATED FOOT MEASUREMENT USING YOLOv8 AND A REFERENCE COIN

Abstract

Correct foot size is vital for medical uses such as diabetic foot screening, orthotics design, sports performance optimization and e-commerce virtual fitting systems. Manual measurement has been a time-consuming, inaccurate and outdated practice for the digital commerce world that loses more than $360 billion in sizing returns each year for retailers. This research suggests an automatic foot measurement system based on the YOLOv8 object detection model, which has state-of-the-art speed (10 ms inference time) and accuracy (0.99 mAP) than the previous object detection models. The system incorporates a reference coin for scale calibration, and is equipped with state-of-the-art image processing algorithms such as adaptive thresholding, Canny edge detection and semantic segmentation to ensure measurement accuracy of 0.05 cm with an average error and 0.02 cm standard deviation. The results were validated with 500 manual measurements and the performance proved to be better (paired t-test, p<0.01). The system has transformative potential across industries, with case studies reporting a 30% decrease in e-commerce returns and a 25% increase in the quality of the fit of prostheses. The development of future mobile application, AR/IoT for real time virtual fitting is planned.