def multiple_input_images(checkpoint, num_styles, input_images_dir,
input_images, which_styles):
"""Added by Raul Gombru. Computes style transfer for a list of images"""
result_images = {}
with tf.Graph().as_default(), tf.Session() as sess:
# Need to initialize the image var to load the model
image_path = input_images_dir + input_images[0]
image = np.expand_dims(
image_utils.load_np_image(os.path.expanduser(image_path)), 0)
stylized_images = model.transform(
tf.concat([image for _ in range(len(which_styles))], 0),
normalizer_params={
'labels': tf.constant(which_styles),
'num_categories': num_styles,
'center': True,
'scale': True
},
reuse=tf.AUTO_REUSE)
# Load the model
_load_checkpoint(sess, checkpoint)
# Stylize images
for image_name in input_images:
image_path = input_images_dir + image_name
image = np.expand_dims(
image_utils.load_np_image(os.path.expanduser(image_path)), 0)
stylized_images = model.transform(tf.concat(
[image for _ in range(len(which_styles))], 0),
normalizer_params={
'labels':
tf.constant(which_styles),
'num_categories':
num_styles,
'center':
True,
'scale':
True
},
reuse=tf.AUTO_REUSE)
stylized_images = stylized_images.eval()
for which, stylized_image in zip(which_styles, stylized_images):
result_images[image_name.split('.')[0] + '_' +
str(which)] = stylized_image[None, ...]
return result_images
def main(unused_argv):
tf.logging.set_verbosity(tf.logging.INFO)
style_files = _parse_style_files(os.path.expanduser(FLAGS.style_files))
with tf.python_io.TFRecordWriter(
os.path.expanduser(FLAGS.output_file)) as writer:
for style_label, style_file in enumerate(style_files):
tf.logging.info(
'Processing style file %s: %s' % (style_label, style_file))
feature = {'label': _int64_feature(style_label)}
style_image = image_utils.load_np_image(style_file)
buf = io.BytesIO()
skimage.io.imsave(buf, style_image, format='JPEG')
buf.seek(0)
feature['image_raw'] = _bytes_feature(buf.getvalue())
if FLAGS.compute_gram_matrices:
with tf.Graph().as_default():
style_end_points = learning.precompute_gram_matrices(
tf.expand_dims(tf.to_float(style_image), 0),
# We use 'pool5' instead of 'fc8' because a) fully-connected
# layers are already too deep in the network to be useful for
# style and b) they're quite expensive to store.
final_endpoint='pool5')
for name in style_end_points:
feature[name] = _float_feature(
style_end_points[name].flatten().tolist())
example = tf.train.Example(features=tf.train.Features(feature=feature))
writer.write(example.SerializeToString())
tf.logging.info('Output TFRecord file is saved at %s' % os.path.expanduser(
FLAGS.output_file))
def main(unused_argv):
tf.logging.set_verbosity(tf.logging.INFO)
style_files = _parse_style_files(os.path.expanduser(FLAGS.style_files))
with tf.python_io.TFRecordWriter(
os.path.expanduser(FLAGS.output_file)) as writer:
for style_label, style_file in enumerate(style_files):
tf.logging.info(
'Processing style file %s: %s' % (style_label, style_file))
feature = {'label': _int64_feature(style_label)}
style_image = image_utils.load_np_image(style_file)
buf = io.BytesIO()
scipy.misc.imsave(buf, style_image, format='JPEG')
buf.seek(0)
feature['image_raw'] = _bytes_feature(buf.getvalue())
if FLAGS.compute_gram_matrices:
with tf.Graph().as_default():
style_end_points = learning.precompute_gram_matrices(
tf.expand_dims(tf.to_float(style_image), 0),
# We use 'pool5' instead of 'fc8' because a) fully-connected
# layers are already too deep in the network to be useful for
# style and b) they're quite expensive to store.
final_endpoint='pool5')
for name, matrix in style_end_points.iteritems():
feature[name] = _float_feature(matrix.flatten().tolist())
example = tf.train.Example(features=tf.train.Features(feature=feature))
writer.write(example.SerializeToString())
tf.logging.info('Output TFRecord file is saved at %s' % os.path.expanduser(
FLAGS.output_file))
def change_style(output):
# without header, It doesn't work
res = requests.get(output['file']['url_private_download'],
headers={'Authorization': 'Bearer %s' % BOT_API},
stream=True)
# TODO: Now this doesn't use the image on memory.
# TODO: Explore better resize.
img_arr = misc.imresize(misc.imread(BytesIO(res.content), mode='RGB'),
(200, 200))
misc.imsave('../imgs/input/temp.jpg', img_arr)
img = np.expand_dims(
image_utils.load_np_image(
os.path.expanduser('../imgs/input/temp.jpg')), 0)
# download models
Style.download_checkpoints('check_points')
style = Style()
generated_imgs = style.generate_image(img)
for i, generated_img in enumerate(generated_imgs):
# TODO: Now this doesn't use the image on memory.
file_name = '../imgs/output/' + 'generated_' + str(i) + '.jpg'
misc.imsave(file_name, generated_img)
slack.files.upload(file_name,
filename=file_name,
channels=output['channel'])
def main(unused_argv=None):
# Load image
image = np.expand_dims(image_utils.load_np_image(FLAGS.input_image), 0)
which_styles = ast.literal_eval(FLAGS.which_styles)
with tf.Graph().as_default(), tf.Session() as sess:
stylized_images = model.transform(
tf.concat(0, [image for _ in range(len(which_styles))]),
normalizer_params={
'labels': tf.constant(which_styles),
'num_categories': FLAGS.num_styles,
'center': True,
'scale': True
})
model_saver = tf.train.Saver(tf.all_variables())
checkpoint = FLAGS.checkpoint
if tf.gfile.IsDirectory(checkpoint):
checkpoint = tf.train.latest_checkpoint(checkpoint)
tf.logging.info(
'loading latest checkpoint file: {}'.format(checkpoint))
model_saver.restore(sess, checkpoint)
stylized_images = stylized_images.eval()
for which, stylized_image in zip(which_styles, stylized_images):
image_utils.save_np_image(
stylized_image[None, ...],
'{}/{}_{}.png'.format(FLAGS.output_dir, FLAGS.output_basename,
which))
def initialize_image():
input_image = './images/me.jpg'
image = np.expand_dims(
image_utils.load_np_image(os.path.expanduser(input_image)), 0)
print(image.shape)
print(type(image[0, 0, 0, 0]))
images['latest'] = image
def main(unused_argv=None):
# Load image
image = np.expand_dims(image_utils.load_np_image(FLAGS.input_image), 0)
output_dir = os.path.expanduser(FLAGS.output_dir)
if not os.path.exists(output_dir):
os.makedirs(output_dir)
which_styles = ast.literal_eval(FLAGS.which_styles)
if isinstance(which_styles, list):
_multiple_images(image, which_styles, output_dir)
elif isinstance(which_styles, dict):
_multiple_styles(image, which_styles, output_dir)
else:
raise ValueError('--which_styles must be either a list of style indexes '
'or a dictionary mapping style indexes to weights.')
def main(unused_argv=None):
# Load image
image = np.expand_dims(image_utils.load_np_image(FLAGS.input_image), 0)
output_dir = os.path.expanduser(FLAGS.output_dir)
if not os.path.exists(output_dir):
os.makedirs(output_dir)
which_styles = ast.literal_eval(FLAGS.which_styles)
if isinstance(which_styles, list):
_multiple_images(image, which_styles, output_dir)
elif isinstance(which_styles, dict):
_multiple_styles(image, which_styles, output_dir)
else:
raise ValueError(
'--which_styles must be either a list of style indexes '
'or a dictionary mapping style indexes to weights.')
def main(unused_argv=None):
#print("begin main...")
style_name = os.environ.get('Http_X_Style_Name', 'varied')
style_index = os.environ.get('Http_X_Style_Index', '1')
which_styles = os.environ.get('Http_X_Which_Styles',
"{%s:1}" % style_index)
#print("Using style: " + style_name)
#print("Using index: " + style_index)
# Monet
# ---------------------
# --num_styles=10 \
# --checkpoint=/magenta-models/multistyle-pastiche-generator-monet.ckpt \
# --input_image=$IMAGE \
# --which_styles="{$i:1}" \
# --output_dir="out_""$IMAGE" \
# --output_basename="monet_styles"
# Varied
# ---------------------
# --num_styles=32 \
# --checkpoint=/magenta-models/multistyle-pastiche-generator-varied.ckpt \
# --input_image=$IMAGE \
# --which_styles="{$i:1}" \
# --output_dir="out_""$IMAGE" \
# --output_basename="varied_styles"
if style_name == 'monet':
FLAGS.num_styles = 10
FLAGS.checkpoint = "/magenta-models/multistyle-pastiche-generator-monet.ckpt"
FLAGS.input_image = input_image_name
FLAGS.which_styles = which_styles
FLAGS.output_dir = "out_content.jpg"
FLAGS.output_basename = "monet_styles"
elif style_name == 'varied':
FLAGS.num_styles = 32
FLAGS.checkpoint = "/magenta-models/multistyle-pastiche-generator-varied.ckpt"
FLAGS.input_image = input_image_name
FLAGS.which_styles = which_styles
FLAGS.output_dir = "out_content.jpg"
FLAGS.output_basename = "varied_styles"
else:
raise ValueError(
'Style %s is not supported. Accepted values are "monet" or "varied"'
% style_name)
#print("loading image...")
# Load image
image = np.expand_dims(
image_utils.load_np_image(os.path.expanduser(FLAGS.input_image)), 0)
output_dir = os.path.expanduser(FLAGS.output_dir)
if not os.path.exists(output_dir):
os.makedirs(output_dir)
which_styles = ast.literal_eval(FLAGS.which_styles)
if isinstance(which_styles, list):
#print("multiple images...")
_multiple_images(image, which_styles, output_dir)
elif isinstance(which_styles, dict):
#print("multiple styles...")
_multiple_styles(image, which_styles, output_dir)
else:
raise ValueError(
'--which_styles must be either a list of style indexes '
'or a dictionary mapping style indexes to weights.')
