Classification and recommendation of food intake in West Africa for healthy diet using Deep Learning

Abstract

A fascinating area with many applications is that of food item recognition from images. Food recognition is becoming more important in our daily lives because it plays a major part in health-related issues. In this study, a method for categorizing food-related photos using convolutional neural networks has been provided. Convolutional neural networks, in contrast to conventional artificial neural networks, are able to estimate the score function directly from picture pixels. A tensor of outputs is generated by a 2D convolution layer's em ployment of a convolution kernel, which is convolved with the l ayer's input. There are numerous such layers, and the results are concatenated in portions to achieve the final tensor of outputs. The data is also processed using the Max-Pooling function, and the features that result from that processing are employed to train the network. The accuracy of the suggested technique again for classes with in FOOD-101 dataset is 85.78 percent.