lettergram

% Create four filters
[hx, hy] = gradient(fspecial('gaussian',[5 5],sigma));
[hx1, hy1] = altOrientFilter1(hx, hy);
[hx2, hy2] = altOrientFilter2(hx, hy);
[hx3, hy3] = altOrientFilter3(hx, hy);

% Run gaussian filters on the image
gx = double(imfilter(img,hx,'replicate', 'conv'));
gy = double(imfilter(img,hy,'replicate', 'conv'));
gx1 = double(imfilter(img,hx1,'replicate', 'conv'));

lettergram

% Convert an image to double for increased precision
img = double(img);
    
% Find two derived gaussians with respect to x and y
[hx, hy] = gradient(fspecial('gaussian',[5 5],sigma));

% Run the filters over the image, generating a filtered image
% Leaves x and y images of edges
gx = double(imfilter(img,hx,'replicate', 'conv'));
gy = double(imfilter(img,hy,'replicate', 'conv'));

lettergram

% Takes the derivative of a 5x5 gaussian, with a sigma    
[hx, hy] = gradient(fspecial('gaussian',[5 5],sigma));

lettergram

% Convert an image to the Lab color space
colorTransform = makecform('srgb2lab');
img = applycform(rgbImg, colorTransform);

% Make it double to improve representation
img = double(img);

% Find x and y derivative of a 9x9 gaussian
[hx, hy] = gradient(fspecial('gaussian',[9 9],sigma));

lettergram

function T = PCA(matrix)
    [U,S,V] = svd(matrix * transpose(matrix));
    T = U * S
end

lettergram

import numpy as np

matrix = np.matrix(matrix) * np.matrix(matrix).transpose() 
U,S,V = np.linalg.svd(matrix) 
T = U * S

lettergram

% Find score matrix for PCA %
% Outputs score between column s and e %
function W = PCA(matrix, s, e)
    [U,S,V] = svd(matrix * transpose(matrix));
    T = U * sqrt(S);
    T = W(:,s:e);
end

lettergram

% Graphs components %

file = 'seeds_dataset.csv';
x = load(file);

% SVD matrix 
[U,S,V] = svd(x * transpose(x));
    
% An eigenvalue is the diagonal matrix of S off the SVD component
eV = diag(sqrt(S));

lettergram

'''
Full code on: https://github.com/lettergram/PCA
read() obtains every image from training data 
http://abel.ee.ucla.edu/cvxopt/_downloads/mnist.py
img = single image from read()
img[0] - id
img[1] - image of digit
'''

# Normalize the image

lettergram

u, sigma, test = train(read(), 30)
right = 0.0
total = 0.0
for img in test:
    matrix = np.subtract(img[1], np.mean(img[1])) # Center
    matrix = np.reshape(matrix, (-1, 1))
    covImg = np.dot(matrix, np.reshape(matrix, (1, -1))).T

    index = 0;
    best = 0.0