Scholarlink

A person’s body mass index (BMI) is a vital indicator of their health. Through extensive research, it was determined that traditional methods of BMI calculation can be time-consuming. We developed a technique that uses deep learning to predict a person’s BMI, age, and gender from facial images. Our system uses Multi-Task Cascaded Convolutional Neural Networks (MTCNN) to detect faces by cropping the face out of an input image and detecting facial landmarks based on the 5-point facial landmark detection algorithm from the input image. The cutting-edge pre-trained models will be finetuned, including VGG-Faces, ResNet (Residual Neural Network), and VGG16 (Visual Geometric Groups with 16 layers) on a public dataset of 1530 prisoners from Polk County Prison. This dataset contains a multifarious range of faces, including different races, ages, and genders. The uploaded CSV file contains the heights, weights, and gender of the training images. After the image is passed into the network, it generates BMI, age, and gender predictions for the input image. This system uses an efficient face recognition mechanism to identify the age, gender, and BMI of a person using a multi-task BMI prediction model.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Deep Learning Approach for Non - invasive Body Mass Index Calculation

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • S. Harish Nandhan,
  • J. Remoon Zean,
  • A. R. Mahi,
  • R. Meena,
  • S. Mahalakshmi

摘要

A person’s body mass index (BMI) is a vital indicator of their health. Through extensive research, it was determined that traditional methods of BMI calculation can be time-consuming. We developed a technique that uses deep learning to predict a person’s BMI, age, and gender from facial images. Our system uses Multi-Task Cascaded Convolutional Neural Networks (MTCNN) to detect faces by cropping the face out of an input image and detecting facial landmarks based on the 5-point facial landmark detection algorithm from the input image. The cutting-edge pre-trained models will be finetuned, including VGG-Faces, ResNet (Residual Neural Network), and VGG16 (Visual Geometric Groups with 16 layers) on a public dataset of 1530 prisoners from Polk County Prison. This dataset contains a multifarious range of faces, including different races, ages, and genders. The uploaded CSV file contains the heights, weights, and gender of the training images. After the image is passed into the network, it generates BMI, age, and gender predictions for the input image. This system uses an efficient face recognition mechanism to identify the age, gender, and BMI of a person using a multi-task BMI prediction model.