Automatic Human Body Measurement Using Computer Vision & Media Pipe

Abstract

Accurate measurements of the human body are important in a variety of settings, including tailoring, fitness tracking, and healthcare. Conventional measurement approaches usually involve human intervention which can be slow, error-likely, and inconvenient. This project describes an approach to automate human body measurements using computer vision methods, the Media pipe framework and the body matrix framework. The system measures body dimensions using images as inputs, and uses Media pipe's pose estimation and tracking as a way to identify key landmarks on the human body. The landmarks are then used in conjunction with computer vision algorithms to accurately calculate the measurements of interest (height, waist circumference, arm length, etc.). The system is also designed to measure accurately using calibration techniques to account for scaling and posture of the person being measured. The system is easy to use, requiring minimal input from the user, and works with readily available mobile phones or other devices. This technology attempts to simplify the current options available in the tailoring and fashion environments by providing a cost-efficient, productive, and easy option to assess body measurements when compared with conventional structure. The project also investigates the possibilities for integration with augmented reality for real-time visualization of measurements and designs to improve usability. The system was to verify that it would perform robustly across different body types, lighting conditions, and environments. The results demonstrate the potential of combining Media pipe's real-time capabilities with some custom computer vision algorithms to create a scalable and practical solution to automated human body measurement.