Face Recognition Using PCA and LDA

Face recognition implementation using PCA and LDA methodologies. The program's primary workflow involves loading image files and applying preprocessing operations. It subsequently calls the face recognition algorithm and finally visualizes the resulting accuracy metrics. The implementation begins with data normalization prior to file reading, ensuring data comparability across different samples. The core algorithmic processing involves executing PCA for dimensionality reduction and feature extraction, followed by LDA for optimal class separation. The program then partitions the dataset into training and testing subsets, utilizing the training set to train a classifier (typically k-NN or SVM). Finally, the system evaluates classifier accuracy using the test set and generates performance visualization charts to provide intuitive insights into the classifier's effectiveness. Key implementation aspects include: - Data normalization using z-score or min-max scaling - PCA implementation through covariance matrix eigen decomposition - LDA application for maximizing inter-class variance while minimizing intra-class variance - Dataset splitting with stratified sampling for balanced class distribution - Accuracy evaluation using confusion matrices and ROC curves - Visualization through matplotlib/seaborn for performance metrics plotting