Indian Food Image Classification with Transfer Learning

Abstract

Image classification has become easier with deep learning and availability of larger datasets and computational resources. The Convolutional neural network is the most popular and widely used image classification technique in the recent days. In this paper image classification is performed on Indian food dataset using different transfer learning techniques. The food plays important role in human's life as it provides us different nutrients and hence it is necessary for every individual to keep a watch on their eating habits. Therefore, food classification is a quintessential thing for a healthier life style. Unlike the traditional methods of building a model from the scratch, pre trained models are used in this project which saves the computation time and cost and also has given better results. The Indian food dataset of 20 classes with 500 images in each class is used for training and validating. The models used are IncceptionV3, VGG16, VGG19 and ResNet. After experimentation it was found that Google InceptionV3 outperformed other models with an accuracy of 87.9% and loss rate of 0.5893.