Group3_ImageClasssification

Nigerian Food Classification Project Documentation

Introduction

The Nigerian Food Classification project aims to develop a deep learning model using PyTorch to classify various Nigerian food dishes. The project utilizes a dataset sourced from Kaggle containing images of different Nigerian foods, including Jollof Rice, Pounded Yam, Egusi Soup, Suya, and others. This documentation provides an overview of the project, including its objectives, dataset, model architecture, training process, and usage instructions.

Objectives

• Develop an image classification model capable of accurately identifying various Nigerian food dishes.
• Utilize deep learning techniques, specifically convolutional neural networks (CNNs), implemented using PyTorch.
• Train the model on a dataset consisting of images of Nigerian foods sourced from Kaggle.
• Evaluate the model's performance using appropriate metrics and techniques.

Dataset

The dataset used in this project consists of images of various Nigerian food dishes collected from Kaggle. The dataset contains a total of X images belonging to Y classes, including:

• Jollof Rice
• Pounded Yam
• Egusi Soup
• Suya
• Other Nigerian food dishes

The images are organized into a directory structure where each class has its folder containing the respective images.

Result and Discussion

Our research shows the performance of MobileNetV2 model on indigenous Nigeria food image classification achieved the best performance performance with an prediction score of 80%. However,we have not explored some others image classification models such as AlexNet, VGGNet,ResNet etc to compare their performance on the Nigeria food image classification Link to the video Presentation:

Recommendation

In this research study, we proposed a framework for Indigenous Nigeria food image classification. An accuracy of 80% was achieved on the second iteration which was an improvement on the 73%, 80% which was recorded at every iterations respectively. The results obtained revealed that the framework is capable of producing state-of the-art results due to a high level of accuracy of the classifications obtained given the relatively limited training dataset used.

Further work should dig further on food image classification according to Nigeria regions such as North, East, south and west types of foods

Team Members

Bala Mairiga Abduljalil | ballaabduljalil@gmail.com
Al-Amin Musa Magaga | alaminmusamagaga@gmail.com
Aminu Hamza Nababa | alaminhnab4@gmail.com
Lurwanu Abdullahi | lurwanabdullahi2107@gmail.com