def get_pca_features(self):
# Get the list of images
img_list = []
y = []
with open(IMAGE_PATH + TRAIN_LIST, 'r') as f:
while True:
filename = f.readline()
# Zero length indicates EOF
if len(filename) == 0:
break
filename = filename.rstrip()
target = filename.split('.')
y.append(target[0])
# get filename without return
img_list.append(IMAGE_PATH + 'train/' + filename)
X = None
# X = np.empty((len(img_list),))
# Read each image for PCA
for row, img_name in zip(range(len(img_list)), img_list):
img = cv2.imread(img_name, cv2.IMREAD_GRAYSCALE)
cv2.imshow('image', img)
# Calculate the lbp histograms
features = lbp.circular_lbp(img, 2, 8)
cv2.imshow('LBP', np.int8(features))
# separate the image into grids and get histograms
sizex, sizey = features.shape
# Set the size
if row is 0:
blocksx = sizex/FaceRec.__block_size[0] + 1
blocksy = sizey/FaceRec.__block_size[1] + 1
X = np.empty((len(img_list), blocksx * blocksy * 256))
hists = []
for i in range(0, sizex, FaceRec.__block_size[0]):
for j in range(0, sizey, FaceRec.__block_size[1]):
hists.append(np.bincount(
features[i:i+FaceRec.__block_size[0],
j:j+FaceRec.__block_size[1]].ravel(),
minlength=256))
X[row, :] = (np.array(hists).ravel())
print X.shape
# Train the Classifier
self.pca.fit(X)
# Storing the data
self.features = np.empty((len(X), 3))
for i in range(len(X)):
self.features[i, :] = self.pca.transform(X[i, :])
self.targets = y
print len(self.features)
def test(self, img):
features = lbp.circular_lbp(img, 2, 8)
hists = []
sizex, sizey = features.shape
X = []
for i in range(0, sizex, FaceRec.__block_size[0]):
for j in range(0, sizey, FaceRec.__block_size[1]):
hists.append(np.bincount(
features[i:i+FaceRec.__block_size[0],
j:j+FaceRec.__block_size[1]].ravel(),
minlength=256))
X.append(np.array(hists).ravel())
X = self.pca.transform(np.array(X))
prediction = self.clf.predict(X)
return prediction
