N = input("Enter N value:")
############################################################################################
#################################### EIGEN FACES TRAINING AND TESTING #####################
############################################################################################
misrate = [[0.0 for x in xrange(3*N+1)] for x in xrange(10)]
total = [0.0 for x in xrange(10)]
percent = 0
for i in xrange(10) :
tr,t = randomgenerator(N)
#print r,test
Xtrain,ytrain = read_images(os.getcwd()+"\..\data", N, tr)
Xtest,ytest = read_images(os.getcwd()+"\..\data", 11-N, t)
# Princical Component Analysis for feature reduction
eigenvectors, Z, mu = pca(vectorize(Xtrain), ytrain)
l = 0
for num_components in range(0,15*N+1,5):
W,M = eigenface(eigenvectors,Z,num_components)
Wnew = test(vectorize(Xtest), ytest, W, M, mu)
count = 0
idx = knnsearch(W,Wnew)
idx = [k+1 for k in idx]
idx.insert(0,0)
count = 0
for j in xrange(1,16) :
for k in xrange(1,12-N) :
if idx[(j-1)*(11-N)+k] < (j-1)*N+1 or idx[(j-1)*(11-N)+k] > j*N :
count = count+1
misrate[i][l] = count/float(15*(11-N))
l = l+1
misrate = [[0.0 for x in xrange(3*N+1)] for x in xrange(10)]
misrate = np.asarray(misrate)
total = [0.0 for x in xrange(10)]
y = []
k = 1
for i in xrange(15):
for j in xrange(N):
y.append(k)
k = k+1
percent = 0
for ei in xrange(10) :
tr,t = randomgenerator(N)
x,yyy = read_images(os.getcwd()+"\..\data", N, tr)
Xtest,ytest = read_images(os.getcwd()+"\..\data", 11-N, t)
x = np.asarray(vectorize(x))
y = np.asarray(y).reshape(-1,1)
Xtest = np.asarray(vectorize(Xtest))
x = x.T
ii = 0
for num_components in xrange(5,15) :
[W,mu] = fisherfaces(x,y,num_components)
#W = np.asarray(W)
w = np.dot(x,W)
wp = np.dot(Xtest.T,W)
idx = ffknnsearch(w,wp)
idx = [k+1 for k in idx]
idx.insert(0,0)
count = 0
for j in xrange(1,16) :
for k in xrange(1,12-N) :
N = input("Enter N value:")
############################################################################################
#################################### EIGEN FACES TRAINING AND TESTING #####################
############################################################################################
misrate = [[0.0 for x in xrange(3 * N + 1)] for x in xrange(10)]
total = [0.0 for x in xrange(10)]
percent = 0
for i in xrange(10):
tr, t = randomgenerator(N)
#print r,test
Xtrain, ytrain = read_images(os.getcwd() + "\..\data", N, tr)
Xtest, ytest = read_images(os.getcwd() + "\..\data", 11 - N, t)
# Princical Component Analysis for feature reduction
eigenvectors, Z, mu = pca(vectorize(Xtrain), ytrain)
l = 0
for num_components in range(0, 15 * N + 1, 5):
W, M = eigenface(eigenvectors, Z, num_components)
Wnew = test(vectorize(Xtest), ytest, W, M, mu)
count = 0
idx = knnsearch(W, Wnew)
idx = [k + 1 for k in idx]
idx.insert(0, 0)
count = 0
for j in xrange(1, 16):
for k in xrange(1, 12 - N):
if idx[(j - 1) * (11 - N) +
k] < (j - 1) * N + 1 or idx[(j - 1) *
(11 - N) + k] > j * N:
count = count + 1