Food Cuisine Classification by Convolutional Neural Network based Transfer Learning Approach

Abstract

Food image classification is considered as a one of the uplift applications of visual food object recognition in the area of food image processing. Deep learning provides great outcomes in various challenging domains with multiple layers to constitute the inattention of data to build computational models. With this success, many studies have put forward deep-learning-based food image classification models and attained better performances collated with conventional machine learning models. We proposed a deep CNN-based food classification method for food identification with transfer learning and the fine-tuning based on the ResNet and InceptionV3 models. Comparisons of both networks are performed with sixteen and three classes of own Indian food image datasets. Inception V3 achieved more accuracy compared to ResNet-50 when more numbers of food image classes are considered.