A study on Food Classification using Convolutional Neural Network

Maduka, Chidiogo M, Elias, Fanuel, Ekpo, Sunday ORCID: https://orcid.org/0000-0001-9219-3759, Landa, Andre, Bakare, Oluwatoyin, Omole, Richard and Kingsley, Aniebiet (2024) A study on Food Classification using Convolutional Neural Network. In: The Third International Adaptive and Sustainable Science, Engineering and Technology Conference (ASSET 2024), 16 July 2024 - 18 July 2024, Manchester, United Kingdom. (In Press)

Accepted Version File not available for download. Available under License In Copyright. Download (1MB)

Abstract

Understanding the food class and its numerous blends has been an interesting area of research as more people become more health conscious of not only how appetizing and savory a meal may appear to be but also its nutritional content. With the advent of social media and ever-increasing availability of data in this regard with heavy research and sophisticated computational components, models and methodologies are leveraging on this. This research paper is mostly driven to contribute to the wealth of knowledge already obtained in this area by looking to use pretrained models to reliably predict or label food images. It is research heavily reliant on comparing the performance of 3 models namely, Alexnet, ResNet18 and Vgg16 in food image classification. Upon training, validating, and testing all 3 models, Resnet18 ranked the topmost performing model based on an accuracy of approximately 88.4% on unseen data while also achieving the least loss values which were based on the evaluation of ground truth labels as against the predicted labels.