Linear Regression vs. Deep Learning: A Simple Yet Effective Baseline for Human Body Measurement.

Area of Science:

• Computer Vision
• Biometrics
• Machine Learning

Background:

• Accurate human body measurement is crucial for various applications, including apparel fitting and health monitoring.
• Current methods often rely on complex equipment like 3D scanners or image analysis, limiting accessibility.
• There is a need for simpler, self-estimable methods for body measurement.

Purpose of the Study:

• To propose a novel linear regression model for estimating human body measurements.
• To evaluate the model's performance against existing state-of-the-art techniques.
• To establish a standardized baseline for reproducible body measurement estimation.

Main Methods:

• Developed a linear regression model utilizing self-estimated inputs such as height and weight.
• Evaluated the model's accuracy against methods using point clouds and images.
• Compared performance against established state-of-the-art and deep learning approaches on public datasets.

Main Results:

• The proposed linear regression model demonstrated comparable performance to state-of-the-art methods.
• The model outperformed several deep learning models on public datasets.
• The simplicity of the model makes it a suitable baseline and convenient for applications like virtual try-on.

Conclusions:

• A simple linear regression model using self-estimated data provides accurate human body measurements.
• The model offers a practical and efficient alternative to complex measurement techniques.
• Guidelines for standardized estimation are provided to enhance result repeatability.