help='Standard deviation for random init.')
parser.add_argument('--gpu', type=str, default='')
parser.add_argument('--allow_growth', default=False, action='store_true')
args = parser.parse_args()
# Arguments parsed
# Split the extension
output_path, ext = os.path.splitext(args.output_path)
if ext == '':
ext = '.png'
config_gpu(args.gpu, args.allow_growth)
## Precomputing the targets for content and style
# Load content and style images
content_image = preprocess_image_scale(args.content_image_path,
img_size=args.img_size)
style_images = [
preprocess_image_scale(img, img_size=args.style_img_size)
for img in args.style_image_path
]
nb_styles = len(style_images)
model = vgg19.VGG19(weights='imagenet', include_top=False)
outputs_dict = dict([(layer.name, layer.output) for layer in model.layers])
content_features = get_content_features(outputs_dict, args.content_layers)
style_features = get_style_features(outputs_dict,
args.style_layers,
norm_by_channels=args.norm_by_channels)
get_content_fun = K.function([model.input], content_features)
pastiche_net = pastiche_model(None,
width_factor=model_args.width_factor,
nb_classes=model_args.nb_classes,
targets=class_targets)
with h5py.File(checkpoint_path + '.h5', 'r') as f:
pastiche_net.load_weights_from_hdf5_group(f['model_weights'])
inputs = [pastiche_net.input, class_targets, K.learning_phase()]
transfer_style = K.function(inputs, [pastiche_net.output])
num_batches = int(np.ceil(model_args.nb_classes / float(args.batch_size)))
for img_name in os.listdir(args.input_path):
print('Processing %s' % img_name)
img = preprocess_image_scale(os.path.join(args.input_path, img_name),
img_size=args.img_size)
imgs = np.repeat(img, model_args.nb_classes, axis=0)
out_name = os.path.splitext(os.path.split(img_name)[-1])[0]
for batch_idx in range(num_batches):
idx = batch_idx * args.batch_size
batch = imgs[idx:idx + args.batch_size]
indices = batch_idx * args.batch_size + np.arange(batch.shape[0])
if args.use_style_name:
names = style_names[idx:idx + args.batch_size]
else:
names = indices
print(' Processing styles %s' % str(names))
get_style_target = K.function([loss_net.input], s_targets)
gm_lists = [[] for l in args.style_layers]
img_list = []
img_size_list = []
# Get style image names or get all images in the directory
if args.style_imgs is None:
args.style_imgs = os.listdir(args.style_dir)
# Check the image sizes
args.style_img_size = std_input_list(args.style_img_size, len(args.style_imgs), 'Image size')
for img_name, img_size in zip(args.style_imgs, args.style_img_size):
try:
print(img_name)
img = preprocess_image_scale(os.path.join(args.style_dir, img_name),
img_size=img_size)
s_targets = get_style_target([img])
for l, t in zip(gm_lists, s_targets):
l.append(t)
img_list.append(os.path.splitext(img_name)[0])
img_size_list.append(img_size)
except IOError as e:
print('Could not open file %s as image.' %img_name)
mtx = []
for l in gm_lists:
mtx.append(np.concatenate(l))
f = h5py.File(args.gram_dataset_path, 'w')
f.attrs['img_names'] = img_list