Car Brand Classification: An Introductory Project to Computer Vision Using Transfer Learning and CNNs

This project aims to introduce beginners with little to no machine learning experience to the fundamentals of computer vision through a hands-on, practical application: developing a multi-class image classification model that can identify the make of a car from an input image using Python and PyTorch. The project will guide students through the process of using a Convolutional Neural Network (CNN) architecture, leveraging transfer learning by applying the ResNet-50 pre-trained model. By the end of this project, students will have a foundational understanding of key concepts in computer vision, practical experience in model deployment, and a unique project to showcase on a resume.

Objectives

• Understand CV Basics: introduce students to concepts and techniques in image processing, feature extraction, and pattern recognition
• Learn CNN Architecture: explain conceputal structure and function of CNNs and their role in image classification
• Implement Transfer Learning: leverage pre-trained architecture (ResNet-50) to expedite training process and improve model accuracy
• Evaluate and Optimize Model: teach students to interpret results and assess model performance using various metrics and techniques

Project Timeline

Date	Topic	Deliverable / Task
09/22	Project Overview + Intro to Computer Vision	Introduce computer vision
09/29	Intro to CNNs	Teach CNNs
10/06	Data Augmentation + Transfer Learning Practice
10/13	---- Fall Study Break ----	None
10/20	Image Preprocessing + Setup	Setup with Kaggle APIs
10/27	Model Development	Work in groups, apply ResNet-50
11/03	Model Development
11/17	Model Development + Interpreting Performance
11/21	Model Development + Expo Prep.
11/21	Final Project Expo