Face recognition with PCA
Eigenface-based face recognition in MATLAB, my first machine learning project.
NIT Rourkela, summer 2017 (summer research internship)
Under the guidance of Dr. Anup Nandy. My first machine learning project: face recognition with principal component analysis (eigenfaces) in MATLAB, on the AT&T Database of Faces (formerly the ORL dataset) of 40 people with 10 grayscale images each.
- Each training image is flattened into a vector and centered by subtracting the mean face.
- To keep the eigendecomposition small, the covariance is computed in the low-dimensional sample space rather than the pixel space, and eigenvectors with significant eigenvalues are kept as eigenfaces.
- Training and test images are projected onto this face space, and a test image is recognized as the training image with the smallest Euclidean distance.