Estimation of protein from the Images of health drink powders using linear regression and deep learning

Abstract

Human power is very important for the development of this world. newlineToday, maintaining human mind and body in healthy condition is the most newlinevital one, but attaining the same is more difficult. Childhood days play the newlinemost important role in the process of maintaining good health. To maintain a newlinehealthy body, the child should consume the required nutrient in proper newlineproportion. Consumption of the required nutrient in proper proportion leads to newlinethe growth of healthy human in the universe. Contradiction to this, some newlinechildren are born with certain disorders such that they couldn t consume all newlinethe nutrient but can consume only some specified nutrient. For such children newlineit is necessary to provide the required nutrient in a correct ratio. Generally, newlinenutrient is classified into two types 1) macronutrients and 2) micronutrients. newlineCarbohydrates, fats, proteins, fiber and water are macronutrients. Vitamins newlineand minerals are micro nutrients. For example, children with special disorders newlinelike maple syrup urine disease can consume only limited amount of protein newlineper day that their body can digest. On the other hand, children with protein newlineenergy malnutrition need to consume more protein. Hence measurement of newlineprotein in food for such children becomes an important task. newlineIn this research, one of the macronutrient protein is estimated from newlinefood images using image processing. In order to perform the research a new newlinedatabase named as image database for protein estimation is created. The newlinedatabase comprises of 990 images which are captured from 9 health drink newlinepowders using an android phone. newlineProtein content in health drink powder is predicted using image newlinefeatures with linear regression with support vector machine and deep newlineconvolutional neural network. newline