Please write in Jupyter notebook only.

Project-2: Face Recognition Using PCA Problem Statement: Given a face image, retrieve the similar images from the given database of face images. Step-1: collect 10 images of faces (training faces) (face images should be centered). Step2: resize the image to 5050. Step3: convert the images to gray scale images. Step-4: for each image, get the pixels values. Now you will have 5050 pixels. Step-5: represent every image I as a vector T. Now for each image you will have n21 vector where n is 50 . Step-6: compute the face vectors i.e. form a matrix that have each image vector in each column and compute the mean face. Display that face. Step-7: Subtract the average face vector from the face vectors. Step-8: Calculate the covariance matrix, which results in nn matrix. Step-9: Calculate the eigenvalues and eigenvectors from the covariance matrix. Step-10: Choose the K best eigenvectors from step-9. Projection of training samples into the Eigenface space: Step-11: multiply each eigenvalues i.e. eigen vectors with the (face vector -average face vector) i.e. step-7 Step-12: graphically display each face with respect to the eigenvalues. Step-13: read the test image and separate the face from the image. If you already have a seplarated face image i.e. image which have a face centered and resized to 5050, you can skip Step-13. Step-14: calculate the feature vector of the test face and subtract it with the average face. Step-15: project the test image on the eigenspace. Step-16: calculate the Euclidean distance (e) it with each eigenface vectors. Step-17: if e l ' from the training set. e is called the distance within face space. The major disadvantage in this approach is that the system can find a face match for a non-facial image too. For example: if you have an image of a car or Mickey Mouse as a test image, this approach might choose a matched face from the training images. Choosing an appropriate threshold value can avoid this misrecognition. In this project, we are performing a facial recognition. A proper facial identification algorithm can avoid this misrecognition. We assume that all images in training and testing are images of someone's face. Proper cropping/clipping of face image from the raw image, centering the face image, size conversion and gray scale representation would reduce the percentage of error in