How udrasht/face-recognition-using-principal-component-analysis works
Project type and purpose
• This is an academic/data-science mini-project implemented in a Jupyter Notebook. Its purpose is to perform face recognition via Principal Component Analysis (PCA) on the AT&T “faces” dataset, generating eigenfaces, projecting new images, and evaluating recognition accuracy.
File structure
• Top-level files:
• README.md – describes PCA steps, dataset, and results.
• code.ipynb – the main implementation: data loading, preprocessing, PCA training, projection, plotting.
• docs/images – example result plots.
• ATnT/ – dataset folder holding 40 subject directories (s1…s40), each containing 10 grayscale PGM images.
• No separate “src,” “frontend,” or “backend” directories; the entire pipeline lives in the notebook.
README for additional insights
• It outlines six conceptual steps: data collection & preprocessing → compute covariance → eigenvectors/eigenvalues → select principal components → project images → compare & evaluate.
• It shows two result images under docs/images.
Detailed guidance
• “Eigenface Basis” (persisted model)
• “Training Projections” (feature database)
• Enclose the entire flow inside a “Jupyter Notebook” boundary to show orchestration.
• Module with key libraries (e.g., NumPy, Matplotlib).