def find_faces_and_label():
img = cv2.imread('images/group.jpg')
faces = face_recognition.face_locations(img)
# faces = get_faces(img, position=True, greyscale=False, check_eyes=False)
tmp = [img[top:bottom, left:right] for top, right, bottom, left in faces]
predictions = predict_cnn(images=tmp)
from facial_expressions import predict_expression_svm
predictions_expressions = predict_expression_svm(tmp)
for face, prediction, prediction_exp in zip(faces, predictions,
predictions_expressions):
top, right, bottom, left = face
label = prediction_exp
cv2.rectangle(img, (left, top), (right, bottom), (0, 255, 0), 3)
cv2.putText(img,
"%s" % int(prediction), (left + 3, bottom - 4),
cv2.FONT_HERSHEY_DUPLEX,
2, (0, 170, 0),
lineType=cv2.LINE_AA)
cv2.imwrite('output.jpg', img)
cv2.imshow('img', cv2.resize(img, None, fx=0.5, fy=0.5))
cv2.waitKey()
cv2.destroyAllWindows()
def extract_faces_from_groups():
group_images = list(
filter(lambda i: i[-3:] in ['jpg'], listdir(group_images_location)))
print(group_images)
i = 1
for group_image_name in group_images:
image = cv2.imread(group_images_location + group_image_name)
faces = get_faces(image, greyscale=False)
for face in faces:
p1, p2 = predict_cnn(face)
cv2.imwrite(
'%s%s-%s.jpg' % (group_image_faces_location, p1, str(i)), face)
i += 1
def extract_faces_from_test_image():
"""
Method to extract faces from single image with the predicted class as the name of the file.
:return:
"""
image = cv2.imread("images/group3.jpg")
faces = face_recognition.face_locations(image)
index = 0
for face in faces:
top, right, bottom, left = face
label = predict_cnn(image[top:bottom, left:right])[0]
cv2.imwrite("test_images/%s-%s.jpg" % (label, index),
image[top:bottom, left:right])
index += 1
def make_confusion_matrix():
"""
Method to create the confusion matrix for the CNN
:return:
"""
# image = cv2.imread("images/group.jpg")
predicted = []
actual = []
# faces = face_recognition.face_locations(image)
#
# for face in faces:
# top, right, bottom, left = face
#
# label = predict_cnn(image[top:bottom, left:right])[0]
# predicted.append(label)
#
# print(label)
# show_image(image[top:bottom, left:right])
# tmp = input()
# actual.append(label if tmp == "" else tmp)
individuals = listdir(face_folders_location + "test/")
for individual in individuals:
test_individual_folder_path = face_folders_location + "test/" + individual + "/"
filenames = listdir(test_individual_folder_path)
for filename in filenames:
image = cv2.imread(test_individual_folder_path + filename)
prediction = predict_cnn(image)[0]
predicted.append(prediction)
actual.append(individual)
print(predicted)
print(actual)
return actual, predicted
text_train = get_cleaned_text('../data/test_text')
text_test = get_cleaned_text('../data/training_text')
w2v_model = Word2Vec(text_train + text_test,
size=vector_size,
window=5,
min_count=3,
workers=4,
iter=5)
return w2v_model
if __name__ == '__main__':
train_variants = pd.read_csv('../data/training_variants')
test_variants = pd.read_csv('../data/test_variants')
n_text = len(train_variants)
#w2v_model = create_w2v()
#w2v_model = Word2Vec.load('w2v_tt200')
w2v_model = Word2Vec.load('w2v_tt50')
fvec = get_feature_vector2('../data/training_text', train_variants,
n_text + 1)
wvec_size = w2v_model.layer1_size
#nn_model = cnn.test(fvec2)
nn_model = cnn.create_model(fvec[:, :, :], train_variants, epochs=10)
fvec_test = get_feature_vector2('../data/test_text', test_variants)
cnn.predict_cnn(nn_model, fvec_test)
def RecogniseFace(image="", feature_type=None, classifier_name=None):
"""
:param image: path to image (relative or absolute)
:param feature_type: HOG or SURF
:param classifier_name: CNN or SVM
:param creative_mode: 0 = False, 1 = True
:return:
"""
# Prepare the inputs
if image is None or classifier_name is None:
print("Must provide an image and classifier name")
return None
classifier_name = classifier_name.upper()
if not os.path.isfile(image):
print("Image file could not be found")
return None
if classifier_name not in ['CNN', 'SVM', 'KNN']:
print("Must provide a supported classified name: CNN or SVM or KNN")
return None
feature_type = feature_type.upper() if feature_type else ""
if classifier_name == 'SVM' and feature_type.upper() not in [
'HOG', 'SURF'
]:
print('Must provide a supported feature type for SVM: HOG or SURF')
return None
if classifier_name == 'KNN' and feature_type.upper() not in [
'HOG', 'SURF'
]:
print('Must provide a supported feature type for KNN: HOG or SURF')
return None
image_raw = cv2.imread(image)
# Inputs are fine, lets start producing the output
if classifier_name == "CNN":
face_positions = face_recognition.face_locations(image_raw)
face_predictions = []
for (top, right, bottom, left) in face_positions:
prediction = int(predict_cnn(image_raw[top:bottom, left:right])[0])
emo = int(
predict_expression_svm([image_raw[top:bottom, left:right]])[0])
x, y = int((left + right) / 2), int((top + bottom) / 2)
face_predictions.append((prediction, x, y, emo))
return face_predictions
elif classifier_name == "SVM":
face_positions = face_recognition.face_locations(image_raw)
results = predict_svm([
image_raw[top:bottom, left:right]
for (top, right, bottom, left) in face_positions
], feature_type)
emo_results = predict_expression_svm([
image_raw[top:bottom, left:right]
for (top, right, bottom, left) in face_positions
])
results_positions = list([
(int(result), int((left + right) / 2), int(
(top + bottom) / 2), int(emo))
for result, (
top, right, bottom,
left), emo in zip(results, face_positions, emo_results)
])
return results_positions
elif classifier_name == "KNN":
face_positions = face_recognition.face_locations(image_raw)
results = predict_knn([
image_raw[top:bottom, left:right]
for (top, right, bottom, left) in face_positions
], feature_type)
emo_results = predict_expression_svm([
image_raw[top:bottom, left:right]
for (top, right, bottom, left) in face_positions
])
results_positions = list([
(int(result), int((left + right) / 2), int(
(top + bottom) / 2), int(emo))
for result, (
top, right, bottom,
left), emo in zip(results, face_positions, emo_results)
])
return results_positions