MobileNetV2-Enhanced Food Recognition for Image-Based Menu Search

Abstract— This seminar paper investigates the use of Convolutional Neural Networks (CNNs) to identify and classify different food products in order to improve culinary experiences. In particular, the article makes use of the MobileNetV2 architecture to provide precise and effective food recognition. The main goal is to properly analyze the dataset, which consists of pictures of different food products, utilizing convolutional neural networks and the MobileNetV2 approach in order to take advantage of the dataset. In this project, we took a journey through several key stages. We started by compiling a dataset of different cuisines and dividing it into validation and training sets. Then, we made sure the images were all the same size and ready for analysis. Once the model was ready, we tested it on some new food images to see how well it could recognize them. A dataset of 34 classes each consist of more than 300 images. The proposed model provides 84% accuracy. This project was all about teaching a computer to understand and recognize different foods.