class EvaluateImages:
def __init__(self):
self.utility = Utility()
self.file_setup = FileSetup()
self.cv_operations = CvOperations()
self.image_processing = ImageProcessing()
self.file_setup.setup_file_structure()
self.date_id = self.utility.get_date_id()
def get_new_region_of_interest_from_image(self):
for root, dirs, files in os.walk(self.file_setup.dir_in):
for filename in files:
if self.utility.ignore_ds_store(filename):
new_roi_from_image = False
if self.utility.is_image_valid(filename):
img = self.cv_operations.read_image(filename, self.file_setup.dir_in)
img = self.cv_operations.resize(img, 600, 450)
new_roi_from_image = self.image_processing.generate_region_of_interest(img, filename)
if(new_roi_from_image != 1):
print('ROI not found, finalizing process')
return new_roi_from_image
else:
return new_roi_from_image
def classify_images(self):
graph1 = tf.Graph()
with graph1.as_default():
session1 = tf.Session()
with session1.as_default():
classifier = load_model(self.file_setup.model_classification)
for root, dirs, files in os.walk(self.file_setup.dir_classify):
for filename in files:
if self.utility.is_image_valid(filename):
img = self.cv_operations.read_image(filename, self.file_setup.dir_classify_outlines, 'gray_scale')
img = self.cv_operations.resize(img, 64, 64)
data = img.reshape(1,64,64,1)
model_out = classifier.predict(data)
return np.argmax(model_out)
def correct_outlines(self, classImg):
if classImg == 0:
self.image_processing.generate_outlined_images(0)
elif classImg == 1:
self.image_processing.generate_outlined_images(1)
else:
self.image_processing.generate_outlined_images(2)
def evaluate_pix2pix(self, classImg):
if classImg == 0:
pix2pix.evaluatePix2pix(self.file_setup.model_cats, 1.25/1)
elif classImg == 1:
pix2pix.evaluatePix2pix(self.file_setup.model_flowers, 1.25/1)
else:
pix2pix.evaluatePix2pix(self.file_setup.model_pokemons, 1.25/1)
def generate_final_image(self, img_name):
merged_img = self.image_processing.create_final_image(img_name)
def main(self):
classImg = -1
there_is_new_roi = self.get_new_region_of_interest_from_image()
if there_is_new_roi:
self.image_processing.generate_outlined_images(1)
classImg = self.classify_images()
print('detected class: ', classImg)
self.correct_outlines(classImg)
self.evaluate_pix2pix(classImg)
self.generate_final_image(self.date_id)
return classImg, self.date_id
