A comprehensive machine learning framework for image classification using HOG and KNN classifier
This project explores image processing and feature extraction using the Histogram of Oriented Gradients (HOG) method, a widely used technique in computer vision. The goal is to extract meaningful features from images, visualize them, and prepare them for further machine learning applications, such as classification and object recognition.
HOG is a feature descriptor that captures the structural and edge information in images. By analysing the intensity gradients and edge orientations, HOG highlights essential patterns and textures that are robust to variations in lighting and noise. The project involves preprocessing the images by resizing and converting them to grayscale, ensuring computational efficiency and uniformity across the dataset.
The extracted HOG features are visualized alongside the original grayscale images to provide insights into the feature extraction process. This visualization demonstrates how HOG captures key structural aspects, such as edges and shapes, making it an effective representation for tasks like classification and object detection.
This report outlines the methodology, implementation, and results of applying HOG for image feature extraction, emphasizing its importance in modern computer vision workflows. The project forms the foundation for machine learning applications by transforming raw images into meaningful feature vectors suitable for analysis and modelling.