def enhance(path, save_dir):
# Initialization.
images = tf.placeholder(tf.float32, [None, None, None, 1], name='images')
model = SRCNN(images)
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
# Load the saved checkpoint.
saver = tf.train.Saver()
checkpoint_dir = ckpt_dir
if load_ckpt(sess, checkpoint_dir, saver):
print('Successfully loaded checkpoint.')
else:
print('Failed to load checkpoint.')
if os.path.isfile(path):
print('Upscaling image', path, '...')
test_data, test_color = prepare_raw(path)
# Generate super-resolutioned image.
conv_out = model.eval({images: test_data})
conv_out = conv_out.squeeze()
result_bw = revert(conv_out)
result = np.zeros([result_bw.shape[0], result_bw.shape[1], 3], dtype=np.uint8)
result[:, :, 0] = result_bw
result[:, :, 1:3] = test_color
result = cv.cvtColor(result, cv.COLOR_YCrCb2RGB)
save_path = os.path.join(save_dir, os.path.basename(path))
scipy.misc.imsave(save_path, result)
# Upscale_single_image(model, path, save_dir)
print('Finished upscaling image', path)
elif os.path.isdir(path):
for root, dirs, files in os.walk(path):
for im_name in files:
img_path = os.path.join(path, im_name)
print('Upscaling image', img_path, '...')
test_data, test_color = prepare_raw(img_path)
# Generate super-resolutioned image.
conv_out = model.eval({images: test_data})
conv_out = conv_out.squeeze()
result_bw = revert(conv_out)
result = np.zeros([result_bw.shape[0], result_bw.shape[1], 3], dtype=np.uint8)
result[:, :, 0] = result_bw
result[:, :, 1:3] = test_color
result = cv.cvtColor(result, cv.COLOR_YCrCb2RGB)
save_path = os.path.join(save_dir, os.path.basename(img_path))
scipy.misc.imsave(save_path, result)
print('Finished upscaling image', img_path)
print('Finished upscaling all images.')
else:
print(' [*] Invalid input path.')
def enhance(x, path, save_dir):
# Initialization.
model = SRCNN(x)
l2_loss = tf.reduce_mean(tf.square(labels - model))
optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(
l2_loss)
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
print('Generating super-resolutioned image...')
# Load the saved checkpoint.
saver = tf.train.Saver()
if load_ckpt(sess, ckpt_dir, saver):
print('Successfully loaded checkpoint.')
else:
print('Failed to load checkpoint.')
if os.path.isfile(path):
print('Upscaling image', path, '...')
img = cv.imread(path)
h, w = img.shape[0], img.shape[1]
num_hor = math.ceil((h * multiplier - size_input) / 21)
num_ver = math.ceil((w * multiplier - size_input) / 21)
test_data, test_color = prepare_raw(path)
# with h5py.File(data_path, 'r') as hf:
# test_data = np.array(hf.get('data'))
# test_color = np.array(hf.get('color'))
# Generate super-resolutioned image.
conv_out = model.eval({images: test_data
}) # Result in patch of size 21x21.
height, width = conv_out.shape[1], conv_out.shape[2]
# print('conv_out has shape:', conv_out.shape)
result = np.zeros([height * num_hor, width * num_ver, 3])
# original = np.zeros([height * num_hor, width * num_ver, 1])
# print('result has shape:', result.shape)
# print('num_hor =', num_hor, 'num_ver =', num_ver)
i, j = 0, 0
for idx, image in enumerate(conv_out):
j = idx // num_ver
i = idx - j * num_ver
# print('idx =', idx, 'i =', i, 'j =', j)
result[j * height:j * height + height,
i * width:i * width + width, 0] = image.squeeze()
# result = result.squeeze()
result = revert(result)
print('Filling color information...')
i, j = 0, 0
for idx, image in enumerate(test_color):
j = idx // num_ver
i = idx - j * num_ver
result[j * height:j * height + height,
i * width:i * width + width, 1:3] = image
result = cv.cvtColor(result, cv.COLOR_YCrCb2RGB)
# cv.imshow('super-resolution', result)
# cv.waitKey(0)
# save_path = os.path.join('./result', 'test_raw.png')
save_path = os.path.join(save_dir, os.path.basename(path))
scipy.misc.imsave(save_path, result)
print('Finished upscaling image', path)
elif os.path.isdir(path):
for root, dirs, files in os.walk(path):
for im_name in files:
img_path = os.path.join(path, im_name)
print('Upscaling image', img_path, '...')
img = cv.imread(img_path)
h, w = img.shape[0], img.shape[1]
num_hor = math.ceil((h * multiplier - size_input) / 21)
num_ver = math.ceil((w * multiplier - size_input) / 21)
test_data, test_color = prepare_raw(img_path)
# Generate super-resolutioned image.
conv_out = model.eval({images: test_data
}) # Result in patch of size 21x21.
height, width = conv_out.shape[1], conv_out.shape[2]
# print('conv_out has shape:', conv_out.shape)
result = np.zeros([height * num_hor, width * num_ver, 3])
# original = np.zeros([height * num_hor, width * num_ver, 1])
# print('result has shape:', result.shape)
# print('num_hor =', num_hor, 'num_ver =', num_ver)
i, j = 0, 0
for idx, image in enumerate(conv_out):
j = idx // num_ver
i = idx - j * num_ver
# print('idx =', idx, 'i =', i, 'j =', j)
result[j * height:j * height + height,
i * width:i * width + width,
0] = image.squeeze()
# result = result.squeeze()
result = revert(result)
print('Filling color information...')
i, j = 0, 0
for idx, image in enumerate(test_color):
j = idx // num_ver
i = idx - j * num_ver
result[j * height:j * height + height,
i * width:i * width + width, 1:3] = image
result = cv.cvtColor(result, cv.COLOR_YCrCb2RGB)
# cv.imshow('super-resolution', result)
# cv.waitKey(0)
# save_path = os.path.join('./result', 'test_raw.png')
save_path = os.path.join(save_dir,
os.path.basename(img_path))
scipy.misc.imsave(save_path, result)
print('Finished upscaling image', img_path)
print('Finished upscaling all images.')
else:
print(' [*] Invalid input path.')
def generate_SR(x, path, save_dir):
# Initialization.
model = SRCNN(x)
l2_loss = tf.reduce_mean(tf.square(labels - model))
optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(
l2_loss)
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
print('Generating super-resolutioned image...')
# Load the saved checkpoint.
saver = tf.train.Saver()
if load_ckpt(sess, ckpt_dir, saver):
print('Successfully loaded checkpoint.')
else:
print('Failed to load checkpoint.')
if os.path.isfile(path):
img = cv.imread(path)
h, w = img.shape[0], img.shape[1]
num_hor = math.ceil((h - size_input) / 21)
num_ver = math.ceil((w - size_input) / 21)
test_data, test_label = prepare_data(path)
print('test_data has shape:', test_data.shape, 'label has shape:',
test_label.shape)
# Generate super-resolutioned image.
conv_out = model.eval({images: test_data
}) # Result in patch of size 21x21.
height, width = conv_out.shape[1], conv_out.shape[2]
# print('conv_out has shape:', conv_out.shape)
result = np.zeros([height * num_hor, width * num_ver, 1])
original = np.zeros([height * num_hor, width * num_ver, 1])
# print('result has shape:', result.shape)
# print('num_hor =', num_hor, 'num_ver =', num_ver)
i, j = 0, 0
for idx, image in enumerate(conv_out):
j = idx // num_ver
i = idx - j * num_ver
print('idx =', idx, 'i =', i, 'j =', j)
result[j * height:j * height + height,
i * width:i * width + width, :] = image
result = result.squeeze()
result = revert(result)
i, j = 0, 0
for idx, image in enumerate(test_label):
j = idx // num_ver
i = idx - j * num_ver
original[j * height:j * height + height,
i * width:i * width + width, :] = image
original = original.squeeze()
size_original = original.shape
bicubic = scipy.misc.imresize(original, 1 / 3, interp='bicubic')
bicubic = scipy.misc.imresize(bicubic,
size_original,
interp='bicubic')
# Display and save the image.
# cv.imshow('original', original)
# cv.waitKey(0)
# cv.imshow('bicubic', bicubic)
# cv.waitKey(0)
# cv.imshow('super-resolution', result)
# cv.waitKey(0)
# save_path = os.path.join('./result', 'test.png')
save_path = os.path.join(save_dir, os.path.basename(path))
scipy.misc.imsave(save_path, result)
# scipy.misc.imsave('./result/original.png', original)
scipy.misc.imsave('./result/bicubic.png', bicubic)
elif os.path.isdir(path):
for root, dirs, files in os.walk(path):
for im_name in files:
img_path = os.path.join(path, im_name)
print('Testing on image', img_path)
img = cv.imread(img_path)
h, w = img.shape[0], img.shape[1]
num_hor = math.ceil((h - size_input) / 21)
num_ver = math.ceil((w - size_input) / 21)
test_data, test_label = prepare_data(img_path)
# Generate super-resolutioned image.
conv_out = model.eval({images: test_data
}) # Result in patch of size 21x21.
height, width = conv_out.shape[1], conv_out.shape[2]
# print('conv_out has shape:', conv_out.shape)
result = np.zeros([height * num_hor, width * num_ver, 1])
original = np.zeros([height * num_hor, width * num_ver, 1])
# print('result has shape:', result.shape)
# print('num_hor =', num_hor, 'num_ver =', num_ver)
i, j = 0, 0
for idx, image in enumerate(conv_out):
j = idx // num_ver
i = idx - j * num_ver
# print('idx =', idx, 'i =', i, 'j =', j)
result[j * height:j * height + height,
i * width:i * width + width, :] = image
result = result.squeeze()
result = revert(result)
i, j = 0, 0
for idx, image in enumerate(test_label):
j = idx // num_ver
i = idx - j * num_ver
original[j * height:j * height + height,
i * width:i * width + width, :] = image
original = original.squeeze()
size_original = original.shape
bicubic = scipy.misc.imresize(original,
1 / 3,
interp='bicubic')
bicubic = scipy.misc.imresize(bicubic,
size_original,
interp='bicubic')
save_path = os.path.join(save_dir,
os.path.basename(img_path))
scipy.misc.imsave(save_path, result)
save_path = save_dir + 'bicubic_' + os.path.basename(
img_path)
scipy.misc.imsave(save_path, bicubic)
print('Finished writing', os.path.basename(img_path))
else:
print(' [*] Invalid input path.')
def generate_SR(path, save_dir):
# Initialization.
images = tf.placeholder(tf.float32, [None, None, None, 1], name='images')
model = SRCNN(images)
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
print('Generating super-resolutioned image...')
# Load the saved checkpoint.
saver = tf.train.Saver()
if load_ckpt(sess, ckpt_dir, saver):
print('Successfully loaded checkpoint.')
else:
print('Failed to load checkpoint.')
if os.path.isfile(path):
test_data, test_label, color = prepare_data(path)
print('test_data has shape:', test_data.shape, 'label has shape:', test_label.shape)
print('color has shape:', color.shape)
# Generate super-resolutioned image.
conv_out = model.eval({images: test_data}) # Result in patch of size 21x21.
conv_out = conv_out.squeeze()
result_bw = revert(conv_out)
result = np.zeros([result_bw.shape[0], result_bw.shape[1], 3], dtype=np.uint8)
result[:, :, 0] = result_bw
result[:, :, 1:3] = color
result = cv.cvtColor(result, cv.COLOR_YCrCb2RGB)
bicubic = scipy.misc.imresize(test_label, 1 / multiplier, interp='bicubic')
bicubic = scipy.misc.imresize(bicubic, multiplier * 1.0, interp='bicubic')
bicubic = cv.cvtColor(bicubic, cv.COLOR_BGR2RGB)
# Save the image.
save_path = os.path.join(save_dir, os.path.basename(path))
scipy.misc.imsave(save_path, result)
bicubic_path = os.path.join(save_dir, 'bicubic_' + os.path.basename(path))
scipy.misc.imsave(bicubic_path, bicubic)
print('Finished testing', path)
elif os.path.isdir(path):
for root, dirs, files in os.walk(path):
for im_name in files:
img_path = os.path.join(path, im_name)
print('Testing on image', img_path)
test_data, test_label, color = prepare_data(img_path)
print('test_data has shape:', test_data.shape, 'label has shape:', test_label.shape)
print('color has shape:', color.shape)
# Generate super-resolutioned image.
conv_out = model.eval({images: test_data}) # Result in patch of size 21x21.
conv_out = conv_out.squeeze()
result_bw = revert(conv_out)
result = np.zeros([result_bw.shape[0], result_bw.shape[1], 3], dtype=np.uint8)
result[:, :, 0] = result_bw
result[:, :, 1:3] = color
result = cv.cvtColor(result, cv.COLOR_YCrCb2RGB)
bicubic = scipy.misc.imresize(test_label, 1 / multiplier, interp='bicubic')
bicubic = scipy.misc.imresize(bicubic, multiplier * 1.0, interp='bicubic')
bicubic = cv.cvtColor(bicubic, cv.COLOR_BGR2RGB)
# img = cv.imread(img_path)
# h, w = img.shape[0], img.shape[1]
# num_hor = math.ceil((h - size_input) / 21)
# num_ver = math.ceil((w - size_input) / 21)
# test_data, test_label = prepare_data(img_path)
# # Generate super-resolutioned image.
# conv_out = model.eval({images: test_data}) # Result in patch of size 21x21.
# height, width = conv_out.shape[1], conv_out.shape[2]
# # print('conv_out has shape:', conv_out.shape)
# result = np.zeros([height * num_hor, width * num_ver, 1])
# original = np.zeros([height * num_hor, width * num_ver, 1])
# # print('result has shape:', result.shape)
# # print('num_hor =', num_hor, 'num_ver =', num_ver)
# i, j = 0, 0
# for idx, image in enumerate(conv_out):
# j = idx // num_ver
# i = idx - j * num_ver
# # print('idx =', idx, 'i =', i, 'j =', j)
# result[j * height : j * height + height, i * width : i * width + width, :] = image
# result = result.squeeze()
# result = revert(result)
# i, j = 0, 0
# for idx, image in enumerate(test_label):
# j = idx // num_ver
# i = idx - j * num_ver
# original[j * height : j * height + height, i * width : i * width + width, :] = image
# original = original.squeeze()
# size_original = original.shape
# bicubic = scipy.misc.imresize(original, 1/3, interp='bicubic')
# bicubic = scipy.misc.imresize(bicubic, size_original, interp='bicubic')
# Save the image.
save_path = os.path.join(save_dir, os.path.basename(img_path))
scipy.misc.imsave(save_path, result)
bicubic_path = os.path.join(save_dir, 'bicubic_' + os.path.basename(img_path))
scipy.misc.imsave(bicubic_path, bicubic)
print('Finished testing', os.path.basename(img_path))
print('Finished testing all images.')
else:
print(' [*] Invalid input path.')