Calorie estimation of real time south Indian food data using Convolutional Neural Network

Due to recent healthcare advancements, food image categorization is a growing subject of research and interest. High carbohydrate content in Indian cuisine is responsible for the raising diabetes rate in the country. There is very less literature present which classifies Asian dishes. The main objective of this paper is to classify south Indian food dishes and to estimate the calories present in the food. The proposed model uses VGG-16 architecture of Convolutional Neural Network and ResNet50 for feature extraction. The data set used in this paper is a collection of real-time images on south Indian food consisting of 11 food dishes. On the training data, data augmentation is performed to increase the model’s inputs. The accuracy of the model is estimated as 71% for classifying the real-time dataset.