A Comparative Analysis of Deep Learning for Vehicle Make-Model-Year Image Classification

This project focuses on developing an image classification model to accurately identify the Make, Model, and Year of vehicles using Convolutional Neural Networks (CNNs). The process involved extensive data preprocessing, model training, and hyperparameter tuning, utilising both PyTorch and TensorFlow frameworks. Despite the steep learning curve and complexities in understanding CNNs and deep learning models, the project successfully leveraged transfer learning techniques with pre-trained models like ResNet50, VGG16, VGG19, etc. A custom web application was developed using Next.js, integrating the final ResNet50 model to provide real-time vehicle classification and pricing information. The application also incorporates fallback mechanisms using OpenAI's GPT-4o vision capabilities and eBay's API for comprehensive user interactions. This project highlights the challenges and rewards of combining advanced machine learning techniques with practical web development to create a functional and user-friendly tool.