NORMA eResearch @NCI Library

Vegan Ingredient Images Classification using Deep Learning and Transfer Learning

Dabekar, Ketaki (2023) Vegan Ingredient Images Classification using Deep Learning and Transfer Learning. Masters thesis, Dublin, National College of Ireland.

[thumbnail of Master of Science]
Preview
PDF (Master of Science)
Download (2MB) | Preview
[thumbnail of Configuration Manual]
Preview
PDF (Configuration Manual)
Download (779kB) | Preview

Abstract

Nowadays, the rising popularity of veganism and plant-based diets is growing fast. Veganism itself is a growing trend with a lot of business potential. Many people are adopting vegan food, so in the market, new ingredients are being introduced. Many people don’t know much about vegan ingredients due to a lack of knowledge about it. The modern food industry offers a diverse range of products. It is difficult to identify vegan ingredients with complex formulations. There is a growing need for advanced technologies that will assist every person in identifying vegan-friendly ingredients in food products. The application of deep learning and transfer learning shows remarkable success in image classification tasks. In this research project, deep learning, and transfer learning techniques will be used to classify vegan ingredient images. There are seven different types of transfer learning techniques available, but in this research, ResNet50, EfficientNet B0, and InceptionV3 are conducted and compared by their testing accuracy, so that an effective and efficient automated system will be built in the future. There are 95 classification classes available within the dataset. A diverse range of images has been collected to train the model well. The dataset is a rich collection of coloured images and textures also contains compositions that are commonly found in vegan food. There are a few classes in the dataset that are imbalanced, so tackle this issue by implementing the data-augmentation technique. According to performance, Inception V3 and ResNet 50 are the top 2 models. The testing accuracy of the Inception V3 model is approximately 89.42%, while the ResNet 50 model shows 81.54% accuracy.

Item Type: Thesis (Masters)
Supervisors:
Name
Email
Anant, Aaloka
UNSPECIFIED
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
H Social Sciences > HF Commerce > Marketing > Consumer Behaviour
H Social Sciences > HD Industries. Land use. Labor > Specific Industries > Food Industry
Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions: School of Computing > Master of Science in Data Analytics
Depositing User: Ciara O'Brien
Date Deposited: 07 May 2025 13:56
Last Modified: 07 May 2025 13:56
URI: https://norma.ncirl.ie/id/eprint/7505

Actions (login required)

View Item View Item