ECSE415

Computer Vision coursework at McGill University

Assignment 1: Image Filtering

• Thresholding
• Denoising, Sharpening
• Edge Detection (Sobel, Laplacian of Gaussian, Canny)

Assignment 2: Feature Extraction

• SIFT feature invariance
• SIFT feature matching
• object detection using HoG

Assignment 3: Segmentation and Motion

• K-Means and GMM implementations from scratch
• normalized graph-cut and mean shift segmentation
• Lucas-Kanade optic flow detection

Assignment 4: Classifiers

• feature extraction using mean pixel intensities and HoG
• classification of CIFAR-10 dataset using SVMs, Random Forests, and a Voting classifier
• experimentation with hyperparameters to optimize performance

Final Project: Facial Recognition

• Acquired a training and testing dataset of images of five test subjects, with a variety of poses, scales, and accessories.
• Built feature vocabularies using SIFT or Harris Corners + HoG or LBP + bag-of-visual-words, or PCA to determine the method with the best performance.
• Created a variety of classifiers to perform facial recognition, tuning hyperparameters to achieve optimal results.