Final Report Edits

Implementation/Discriminator_Images.py

@@ -0,0 +1,41 @@
+import matplotlib.pyplot as plt
+import os
+import glob
+
+INPUT_DIR = "Results/Discriminator"
+
+
+def main():
+	folders = ["Input", "PixelGAN", "PatchGAN1", "PatchGAN2", "PatchGAN3"]
+
+	plt.figure(figsize=(20, 15))
+	type = 0
+	fsize = 35
+	for file in folders:
+		test_images = sorted(glob.glob(os.path.join(INPUT_DIR, file, '*.jpg')))
+		type += 1
+		counter = 0
+		for img_path in test_images[0:5]:
+			img = plt.imread(img_path)
+			idx = 5 * counter + type
+			plt.subplot(5, 5, idx)
+			plt.imshow(img)
+			if idx == 1:
+				plt.title("Input", fontsize=fsize)
+			elif idx == 2:
+				plt.title("1x1", fontsize=fsize)
+			elif idx == 3:
+				plt.title("16x16", fontsize=fsize)
+			elif idx == 4:
+				plt.title("70x70", fontsize=fsize)
+			elif idx == 5:
+				plt.title("256x256", fontsize=fsize)
+			plt.axis('off')
+			counter += 1
+	# plt.savefig(fname="Artistic_Style3.png", bbox_inches='tight')
+	plt.show()
+	print('Saved Images')
+
+
+if __name__ == '__main__':
+	main()

Implementation/Literature Comparison.py

@@ -0,0 +1,102 @@
+import matplotlib.pyplot as plt
+import tensorflow as tf
+
+INPUT_DIR = "Results/Literature Comparison/"
+
+
+def preprocess_image(image_path):
+	image = plt.imread(image_path)
+
+	# resize to 286x286
+	Ht = tf.shape(image)[0]
+	Wt = tf.cast(tf.math.multiply(1.5, tf.cast(Ht, tf.float16)), tf.int32)
+	image = tf.image.resize(image, [Ht, Wt], method=tf.image.ResizeMethod.NEAREST_NEIGHBOR)
+
+	return image
+
+
+def main():
+
+	plt.figure(figsize=(45, 15))
+	fsize =45
+	img_path = INPUT_DIR
+
+	# Chicago
+	img = preprocess_image(img_path + "chicago.jpg")
+	plt.subplot(2, 5, 1)
+	plt.imshow(img)
+	plt.title("Input", fontsize=fsize)
+	plt.axis('off')
+
+	# Style Image
+	img = preprocess_image(img_path + "Starry_night.jpg")
+	plt.subplot(2, 5, 2)
+	plt.imshow(img)
+	plt.title("Starry Night", fontsize=fsize)
+	plt.axis('off')
+
+
+	# Gatys Starry
+	img = preprocess_image(img_path + "Gatys_Starry.jpg")
+	plt.subplot(2, 5, 3)
+	plt.imshow(img)
+	plt.title("Gatys et al.", fontsize=fsize)
+	plt.axis('off')
+
+
+	# Justin Starry
+	img = preprocess_image(img_path + "Justin_Starry.jpg")
+	plt.subplot(2, 5, 4)
+	plt.imshow(img)
+	plt.title("Johnson et al.", fontsize=fsize)
+	plt.axis('off')
+
+	# CycleGAN Starry
+	img = preprocess_image(img_path + "CycleGAN_Starry.jpg")
+	plt.subplot(2, 5, 5)
+	plt.imshow(img)
+	plt.title("CycleGAN", fontsize=fsize)
+	plt.axis('off')
+
+	# Chicago
+	img = preprocess_image(img_path + "chicago.jpg")
+	plt.subplot(2, 5, 6)
+	plt.imshow(img)
+	plt.title("Input", fontsize=fsize)
+	plt.axis('off')
+
+	# Style Image
+	img = preprocess_image(img_path + "Wave.jpg")
+	plt.subplot(2, 5, 7)
+	plt.imshow(img)
+	plt.title("Gatys et al.", fontsize=fsize)
+	plt.axis('off')
+
+	# Gatys Starry
+	img = preprocess_image(img_path + "Gatys_Wave.jpg")
+	plt.subplot(2, 5, 8)
+	plt.imshow(img)
+	plt.title("Gatys et al.", fontsize=fsize)
+	plt.axis('off')
+
+	# Justin Starry
+	img = preprocess_image(img_path + "Justin_Wave.jpg")
+	plt.subplot(2, 5, 9)
+	plt.imshow(img)
+	plt.title("Johnson et al.", fontsize=fsize)
+	plt.axis('off')
+
+	# CycleGAN Starry
+	img = preprocess_image(img_path + "CycleGAN_Wave.jpg")
+	plt.subplot(2, 5, 10)
+	plt.imshow(img)
+	plt.title("CycleGAN", fontsize=fsize)
+	plt.axis('off')
+
+	# plt.savefig(fname="Artistic_Style3.png", bbox_inches='tight')
+	plt.show()
+	print('Saved Images')
+
+
+if __name__ == '__main__':
+	main()

Implementation/evaluate.py

@@ -0,0 +1,63 @@
+def make_dataset(filepath, method):
+	raw_dataset = tf.data.TFRecordDataset(filepath)
+
+	image_feature_description = {
+		'image/height': tf.io.FixedLenFeature([], tf.int64),
+		'image/width': tf.io.FixedLenFeature([], tf.int64),
+		'image/format': tf.io.FixedLenFeature([], tf.string),
+		'image/encoded': tf.io.FixedLenFeature([], tf.string),
+	}
+
+	def preprocess_image(encoded_image):
+		image = tf.image.decode_jpeg(encoded_image, 3)
+
+		# resize to 286x286
+		# image = tf.image.random_crop(image, [tf.shape(image)[0], tf.shape(image)[0], tf.shape(image)[-1]])
+		image = tf.image.resize(image, [286, 286], method=method)
+
+		# random crop a 256x256 area
+		image = tf.image.random_crop(image, [256, 256, tf.shape(image)[-1]], seed=tf.random.set_seed(1))
+		image = tf.cast(image, tf.float32)
+
+		# normalize from 0-255 to -1 to +1
+		image = tf.divide(image, 127.5) - 1.0
+
+		return image
+
+	def parse_image_function(example_proto):
+		# Parse the input tf.Example proto using the dictionary above.
+		features = tf.io.parse_single_example(example_proto, image_feature_description)
+		encoded_image = features['image/encoded']
+		image = preprocess_image(encoded_image)
+		return image
+
+	parsed_image_dataset = raw_dataset.map(parse_image_function)
+	return parsed_image_dataset
+
+
+def generate_images(model, test_input, name, save=False):
+	prediction = model(test_input)
+	plt.figure(figsize=(15, 15))
+
+	display_list = [test_input[0], prediction[0]]
+	title = ['Input Image', 'Predicted Image']
+
+	for i in range(2):
+		plt.subplot(1, 2, i + 1)
+		plt.title(title[i])
+		# getting the pixel values between [0, 1] to plot it.
+
+		img = display_list[i]
+		Ht = tf.shape(img)[0]
+		Wt = tf.cast(tf.math.multiply(1.25, tf.cast(Ht, tf.float16)), tf.int32)
+		c = tf.shape(img)[-1]
+		img = tf.image.resize(img, [Ht, Wt], method=tf.image.ResizeMethod.NEAREST_NEIGHBOR)
+		if i == 1:
+			img = tf.image.adjust_jpeg_quality(img, jpeg_quality=100)
+
+		plt.imshow(img * 0.5 + 0.5)
+		plt.axis('off')
+
+	if save:
+		plt.savefig(fname=name, bbox_inches='tight')
+	plt.show()

Implementation/generate_figures.py

@@ -0,0 +1,41 @@
+import matplotlib.pyplot as plt
+import os
+import glob
+
+OUTPUT_DIR = "Results/"
+INPUT_DIR = "Results/"
+
+def main():
+	folders = ["Input", "MonetPaintings", "CezannePaintings", "Ukiyo_ePaintings", "VanGoghPaintings"]
+
+	plt.figure(figsize=(20, 31))
+	type = 0
+	fsize = 40
+	for file in folders:
+		test_images = sorted(glob.glob(os.path.join(INPUT_DIR, file, '*.jpg')))
+		type += 1
+		counter = 0
+		for img_path in test_images[20:30]:
+			img = plt.imread(img_path)
+			idx = 5*counter + type
+			plt.subplot(10, 5, idx)
+			plt.imshow(img)
+			if idx == 1:
+				plt.title("Input", fontsize=fsize)
+			elif idx == 2:
+				plt.title("Monet", fontsize=fsize)
+			elif idx == 3:
+				plt.title("Cezanne", fontsize=fsize)
+			elif idx == 4:
+				plt.title("Ukiyo-e", fontsize=fsize)
+			elif idx == 5:
+				plt.title("Van Gogh", fontsize=fsize)
+			plt.axis('off')
+			counter += 1
+	plt.savefig(fname="Artistic_Style3.png", bbox_inches='tight')
+	plt.show()
+	print('Saved Images')
+
+
+if __name__ == '__main__':
+	main()

Implementation/generate_real_pictures.py

@@ -0,0 +1,71 @@
+import matplotlib.pyplot as plt
+import os
+import glob
+
+INPUT_DIR = "Results/"
+
+def main():
+	folders = ["MonetToLandscapeInput", "MonetToLandscape"]
+
+	# plt.figure(figsize=(13, 25))
+	# type = 0
+	# fsize = 30
+	# for file in folders:
+	# 	test_images = sorted(glob.glob(os.path.join(INPUT_DIR, file, '*.jpg')))
+	# 	type += 1
+	# 	counter = 0
+	# 	for img_path in test_images[:6]:
+	# 		img = plt.imread(img_path)
+	# 		idx = 2 * counter + type
+	# 		plt.subplot(6, 2, idx)
+	# 		plt.imshow(img)
+	# 		if idx == 1:
+	# 			plt.title("Input", fontsize=fsize)
+	# 		elif idx == 2:
+	# 			plt.title("Output", fontsize=fsize)
+	# 		plt.axis('off')
+	# 		counter += 1
+	#
+	# plt.savefig(fname="MonetToLandscape1.png", bbox_inches='tight')
+	# plt.show()
+	# print('Saved Images 1')
+	#
+	# plt.figure(figsize=(13, 25))
+	# type = 0
+	# for file in folders:
+	# 	test_images = sorted(glob.glob(os.path.join(INPUT_DIR, file, '*.jpg')))
+	# 	type += 1
+	# 	counter = 0
+	# 	for img_path in test_images[6:12]:
+	# 		img = plt.imread(img_path)
+	# 		idx = 2 * counter + type
+	# 		plt.subplot(6, 2, idx)
+	# 		plt.imshow(img)
+	# 		if idx == 1:
+	# 			plt.title("Input", fontsize=fsize)
+	# 		elif idx == 2:
+	# 			plt.title("Output", fontsize=fsize)
+	# 		plt.axis('off')
+	# 		counter += 1
+	#
+	# plt.savefig(fname="MonetToLandscape2.png", bbox_inches='tight')
+	# plt.show()
+	# print('Saved Images 2')
+
+	plt.figure(figsize=(27, 25))
+	counter = 0
+	folders = ["MonetToLandscape1.png", "MonetToLandscape2.png"]
+	for img_path in folders:
+		img = plt.imread(img_path)
+		counter += 1
+		plt.subplot(1, 2, counter)
+		plt.imshow(img)
+		plt.axis('off')
+
+	plt.savefig(fname="MonetToLandscape.png", bbox_inches='tight')
+	plt.show()
+	print('Saved Images')
+
+
+if __name__ == '__main__':
+	main()