About the dataset:
Labeled Faces in the Wild dataset consists of face photographs designed for studying the problem of unconstrained face recognition. The original dataset contains more than 13,000 images of faces collected from the web.
Agenda:
- In this programming challenge, you will be performing face recognition on the Labeled Faces in the Wild dataset using PyTorch.
- First, you will do Principal Component Analysis (PCA) on the image dataset. PCA is used for dimentionality reduction which is a type of unsupervised learning.
- You will be applying PCA on the dataset to extract the principal components (Top
$k$ eigenvalues). - As you will see eventually, the reconstruction of faces from these eigenvalues will give us the eigen-faces which are the most representative features of most of the images in the dataset.
- Finally, you will train a simple PyTorch Neural Network model on the modified image dataset.
- This trained model will be used prediction and evaluation on a test set.