Fast Food Image Recognition using Transfer Learning

Abstract

Food recognition is a relatively difficult task when compared to traditional image recognition due to the close similarities between different categories of food. We tackle this problem using a Convoluted Neural Network model with and without weights that are pre trained on a much larger dataset. This allows us to utilize a much smaller dataset to fine-tune the weights in order to achieve a higher accuracy in food image recognition. We have compared the accuracy of different Convoluted Neural Network (i.e. VGG16 and AlexNet) models with and without the incorporation of Transfer Learning to correctly classify Fast Food images.