class AstrolabMain:
def __init__(self):
self.neural_network = None
self.image_processor = AstrolabImageProcessor()
self.data_wrapper = None
def start(self):
print("Welcome to Astrolab")
while(1):
try:
print("\nMenu")
print(" 1 - Denoise")
print(" 2 - Train")
print(" 3 - Classify")
print(" 4 - Exit")
user_input = input("Choose an option (numeric): ")
if user_input.isdigit():
action = int(user_input)
if action == 1:
self.denoise()
elif action == 2:
self.train()
elif action == 3:
self.classify()
elif action == 4:
break
else:
print("Invalid choice")
except:
print("\nA error happen:\n" + traceback.format_exc())
def denoise(self):
print("\nDenoising")
image_path = self.read_value(self.is_folder_or_file,"Inform image(s) path: ", "Invalid image(s) path")
image_size = self.read_value(self.is_integer, "Inform image(s) size: ", "Invalid image(s) size")
output_path = self.read_value(self.is_folder, "Inform the output folder: ", "Invalid path")
file_name_list = None
if self.is_file(image_path):
image_path_parts = image_path.split('/')
file_name_list = [image_path_parts[-1]]
image_path_parts.pop()
image_path = '/'.join(image_path_parts)
elif self.is_folder(image_path):
file_name_list = sorted(os.listdir(image_path))
for file_name in file_name_list:
img = cv2.imread(os.path.join(image_path, file_name), cv2.IMREAD_UNCHANGED)
img = self.image_processor.denoise(img, int(image_size))
if img.size != 0:
cv2.imwrite(os.path.join(output_path, file_name), img)
print("Denoising finished")
def train(self):
print("\nTraining")
train_processed_images = self.read_value(self.is_folder,"Inform train images path: ", "Invalid images path")
train_label_file = self.read_value(self.is_file, "Inform train labels path: ", "Invalid label file path")
test_processed_images = self.read_value(self.is_folder, "Inform test images path: ", "Invalid images path")
test_label_file = self.read_value(self.is_file, "Inform test labels path: ", "Invalid label file path")
class_size = int(self.read_value(self.is_integer, "Inform the class count: ", "Invalid class count"))
batch_size = int(self.read_value(self.is_integer, "Inform the batch size: ", "Invalid batch size"))
iterations = int(self.read_value(self.is_integer, "Inform the iterations number: ", "Invalid iterations size"))
if not self.neural_network:
image_size = int(self.read_value(self.is_integer, "Inform the image sizes: ", "Invalid image size"))
image_channels = int(self.read_value(self.is_integer, "Inform the image channels: ", "Invalid image channels"))
self.neural_network = AstrolabNeuralNetwork(image_size, image_channels, class_size)
else:
image_size = self.neural_network.input_shape
image_channels = self.neural_network.input_channels
data = DataWrapper()
data.train.load_labels(train_label_file, class_size)
data.train.load_images_names(train_processed_images, image_size)
data.test.load_labels(test_label_file, class_size)
data.test.load_images_names(test_processed_images, image_size)
self.neural_network.train(data, batch_size, iterations)
print("Training finished")
def classify(self):
print("\nClassifying")
image_path = self.read_value(self.is_file, "Inform image path: ", "Invalid path")
if not self.neural_network:
image_size = int(self.read_value(self.is_integer, "Inform the image sizes: ", "Invalid image size"))
image_channels = int(self.read_value(self.is_integer, "Inform the image channels: ", "Invalid image channels"))
class_size = int(self.read_value(self.is_integer, "Inform the class count: ", "Invalid class count"))
self.neural_network = AstrolabNeuralNetwork(image_size, image_channels, class_size)
image = cv2.imread(image_path)
image = cv2.normalize(image.astype('float'), None, 0.0, 1.0, cv2.NORM_MINMAX)
image_reshaped = np.reshape([image], (-1, self.neural_network.input_dimension))
classification = self.neural_network.classify(image_reshaped)
print("\nPrediction: " + str(np.argmax(classification[0])))
image = cv2.imread(image_path)
image = cv2.resize(image, (400,400), interpolation = cv2.INTER_CUBIC)
cv2.startWindowThread()
cv2.namedWindow("preview")
cv2.imshow("preview", image)
def read_value(self, valid_function, message, error_message):
value = input(message)
while not valid_function(value):
print(error_message)
value = input(message)
return value
def is_file(self, path):
return os.path.isfile(path)
def is_folder(self, path):
return os.path.isdir(path)
def is_folder_or_file(self, path):
return os.path.isfile(path) or os.path.isdir(path)
def is_integer(self, text):
return text.isdigit()
# img = cv2.imread(os.path.join(train_raw_images, file_name), cv2.IMREAD_UNCHANGED)
# img = image_processor.denoise(img, image_size)
# if img.size != 0:
# cv2.imwrite(os.path.join(train_processed_images, file_name), img)
#
# file_name_list_test = sorted(os.listdir(test_raw_images))
#
# for file_name in file_name_list_test:
# img = cv2.imread(os.path.join(test_raw_images, file_name), cv2.IMREAD_UNCHANGED)
# img = image_processor.denoise(img, image_size)
# if img.size != 0:
# cv2.imwrite(os.path.join(test_processed_images, file_name), img)
#
img = cv2.imread("/media/willgluck/a2aa6a5f-a88a-45c7-af45-d38ccf2b7639/work/DATASET/1581.jpg", cv2.IMREAD_UNCHANGED)
img = image_processor.denoise(img, image_size)
# data = input_data.read_data_sets('MNIST_data', one_hot=True)
# data = DataWrapper()
# data.train.load_labels(train_label_file, 2)
# data.train.load_images_names(train_processed_images, image_size)
# data.test.load_labels(test_label_file, 2)
# data.test.load_images_names(test_processed_images, image_size)
# pre_type_train = 0
# late_type_train = 0
#
# for i in range (0, len(data.train.images_names)):
# batch = data.train.next_batch(1)
# if batch[1][0][1] == 1: