def ReadImages(ListName, FolderName, Label):
#global NumberList
global responseData
global trainData
#global hog
#global cv2
#global imutils
#global winSize
global testData
#global os
global feature
for image in ListName:
face_cascade = cv2.CascadeClassifier(
'haarcascade_frontalface_default.xml')
img = cv2.imread(join(FolderName, image))
imgray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
face = face_cascade.detectMultiScale(imgray)
if len(face) > 0:
feature = HOG(
cv2.resize(
imgray[face[0][1]:face[0][1] + face[0][3],
face[0][0]:face[0][0] + face[0][2]], (100, 100)))
trainData.append(feature.T)
responseData.append(Label)
def ReadImages(ListName, FolderName):
for image in ListName:
# face_cascade =cv2.CascadeClassifier('haarcascade_frontalface_default.xml')
# img = cv2.imread(join(FolderName,image))
# imgray = cv2.cvtColor(img,cv2.COLOR_RGB2GRAY)
# face = face_cascade.detectMultiScale(imgray,minSize = (100, 100),scaleFactor = 1.25) #,scaleFactor = 1.15, minNeighbors = 5,
# print(face)
# if len(face)>0:
# x, y, w, h = face[0]
# cv2.rectangle(imgray, (x, y), (x + w, y + h), (0, 255, 0), 2)
# else:
# pass
# cv2.imshow("Objects found", imgray)
# cv2.waitKey(100)
# sleep(0.3)
# cv2.destroyAllWindows()
# cv2.waitKey(1)
# cv2.waitKey(1)
# cv2.waitKey(1)
# cv2.waitKey(1) #bug in openCV linux
try:
image = cv2.imread(join(FolderName, image))
imgray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
face_cascade = cv2.CascadeClassifier(
'haarcascade_frontalface_default.xml')
face = face_cascade.detectMultiScale(imgray,
minSize=(50, 50),
scaleFactor=1.25)
if len(face) > 0:
feature, hog_image = HOG(cv2.resize(
imgray[face[0][1]:face[0][1] + face[0][3],
face[0][0]:face[0][0] + face[0][2]], (100, 100)),
cells_per_block=(1, 1),
visualise=True)
cropped_image = imgray[face[0][1]:face[0][1] + face[0][3],
face[0][0]:face[0][0] + face[0][2]]
fig, (ax1, ax2) = plt.subplots(1, 2)
ax1.axis('off')
ax1.imshow(cropped_image, cmap=plt.cm.gray)
ax1.set_title('Input image')
ax1.set_adjustable('box-forced')
# Rescale histogram for better display
#hog_image_rescaled = exposure.rescale_intensity(hog_image, in_range=(0, 0.02))
ax2.axis('off')
ax2.imshow(hog_image, cmap=plt.cm.gray)
ax2.set_title('Histogram of Oriented Gradients')
ax1.set_adjustable('box-forced')
plt.show()
else:
continue
except Exception as e:
print(e)
pass
def ReadImages(ListName, FolderName, Label):
global NumberList
global responseData
global testData
global hog
global cv2
global imutils
global winSize
ListName = ListName[0:100]
for image in ListName:
img = cv2.imread(join(FolderName, image))
NumberList.append(img)
feature = HOG(cv2.cvtColor(img, cv2.COLOR_RGB2GRAY))
testData.append(feature.T)
realResponse.append(Label)
def ReadImages(ListName,FolderName,Label):
global NumberList
global responseData
global trainData
global hog
global cv2
global imutils
global winSize
for image in ListName:
img = cv2.imread(join(FolderName,image))
img = cv2.resize(img,(50,50))
NumberList.append(img)
feature = HOG(cv2.cvtColor(img,cv2.COLOR_RGB2GRAY))
trainData.append(feature.T)
responseData.append(Label)
def ReadImages(self, ListName, FolderName, Label):
for image in ListName:
img = cv2.imread(join(FolderName, image))
try:
imgray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
face = self.face_cascade.detectMultiScale(imgray,
minSize=(50, 50),
scaleFactor=1.25)
if len(face) > 0:
feature = HOG(
cv2.resize(
imgray[face[0][1]:face[0][1] + face[0][3],
face[0][0]:face[0][0] + face[0][2]],
(100, 100)))
self.trainData.append(feature.T)
self.responseData.append(Label)
except Exception as e:
print(e)
pass
def inputTestData(self, name):
name = self.trainFolder + str(name) + '/'
images = [f for f in listdir(name) if isfile(join(name, f))]
for image in images:
img = cv2.imread(join(name, image))
try:
imgray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
face = self.face_cascade.detectMultiScale(imgray,
minSize=(50, 50),
scaleFactor=1.25)
if len(face) > 0:
feature = HOG(
cv2.resize(
imgray[face[0][1]:face[0][1] + face[0][3],
face[0][0]:face[0][0] + face[0][2]],
(100, 100)))
self.testData.append(feature.T)
self.testDict[image] = feature.T
except Exception as e:
print(e)
pass
ReadImages(GalImages, Gal, 6)
ReadImages(SandbergImages, Sandberg, 7)
ReadImages(SainaImages, Saina, 8)
svm = NuSVC()
nu_options = np.arange(0.1, 1)
kernel_options = ['linear', 'rbf']
param_grid = dict(kernel=kernel_options, nu=nu_options)
gridSVM = GridSearchCV(svm, param_grid, scoring='accuracy')
X = np.float32(trainData)
y = np.float32(responseData)
gridSVM.fit(X, y)
print(gridSVM.best_score_)
print(gridSVM.best_estimator_)
face_cascade = cv2.CascadeClassifier('haarcascade_frontalface_default.xml')
TestImages = [f for f in listdir(Test) if isfile(join(Test, f))]
for image in TestImages:
img = cv2.imread(join(Test, image))
imgray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
face = face_cascade.detectMultiScale(imgray)
if len(face) > 0:
feature = HOG(
cv2.resize(
imgray[face[0][1]:face[0][1] + face[0][3],
face[0][0]:face[0][0] + face[0][2]], (100, 100)))
pred = gridSVM.predict(feature)
plt.figure()
plt.imshow(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
plt.title(num2name(pred))
pred = np.argmax(pred, axis=1)
plt.figure()
plt.imshow(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
plt.title(predDict[int(pred)])
try:
image = cv2.imread(join(Test, image))
imgray = cv2.cvtColor(image, cv2.COLOR_RGB2GRAY)
face_cascade = cv2.CascadeClassifier(
'haarcascade_frontalface_default.xml')
face = face_cascade.detectMultiScale(imgray,
minSize=(50, 50),
scaleFactor=1.25)
if len(face) > 0:
feature, hog_image = HOG(cv2.resize(
imgray[face[0][1]:face[0][1] + face[0][3],
face[0][0]:face[0][0] + face[0][2]], (100, 100)),
cells_per_block=(1, 1),
visualise=True)
cropped_image = imgray[face[0][1]:face[0][1] + face[0][3],
face[0][0]:face[0][0] + face[0][2]]
fig, (ax1, ax2) = plt.subplots(1, 2)
ax1.axis('off')
ax1.imshow(cropped_image, cmap=plt.cm.gray)
ax1.set_title('Input image')
ax1.set_adjustable('box-forced')
# Rescale histogram for better display
#hog_image_rescaled = exposure.rescale_intensity(hog_image, in_range=(0, 0.02))
ax2.axis('off')
responseData.append(Label)
ReadImages(ZeroImages, Zero, 0)
ReadImages(OneImages, One, 1)
ReadImages(TwoImages, Two, 2)
ReadImages(ThreeImages, Three, 3)
ReadImages(FourImages, Four, 4)
ReadImages(FiveImages, Five, 5)
ReadImages(SixImages, Six, 6)
ReadImages(SevenImages, Seven, 7)
ReadImages(EightImages, Eight, 8)
ReadImages(NineImages, Nine, 9)
trainNp = np.float32(trainData)
responseNp = np.float32(responseData)
X = np.float32(trainNp)
y = np.float32(responseNp)
svm = NuSVC(kernel='linear', nu=0.2)
svm.fit(X, y)
for i in range(len(TestImages)):
img = cv2.imread(join(testFolder, TestImages[i]))
img2 = cv2.resize(img, (28, 28))
featuretest = HOG(cv2.cvtColor(img2, cv2.COLOR_RGB2GRAY))
ytest = svm.predict(featuretest)
plt.figure()
plt.imshow(img)
plt.title(str(ytest))