Approximate estimation of the nutritions of consumed food by deep learning

Deep learning has become a popular sub-study in machine learning in the recent years due to the conclusions of development in computer hardware systems and big data. Deep learning is used successfully in image classification, speech recognition, natural language processing, and so on. Controlled intake of nutrition is recommended as a condition for being a healthy individual. Knowing and monitoring how much food is consumed during the day, following the calorie and nutrition of these foods helps to control healthy nutrition. In this study, an attempt was made to approximate the nutritions of the food at the image level using the Food-pics dataset that contain nutrient images. Convolutional Neural Networks (CNN), a deep learning approach that has been used successfully in image recognition and classification tasks, has been trained with nutrition image training data, and a high classification success value has been achieved. The arithmetic average nutritional values of the five highest predicted nutritional results obtained from the net were taken as the approximate nutrient content.