def save_action(path,action):
action_path=path+str(action)+"/"
utils.make_dir(action_path)
for i,img in enumerate(action.images):
img_path=action_path+str(action)+"_"+str(i)
img=np.reshape(img,(80,40))
utils.save_img(img_path,img)
def save(self,path,name):
for proj,postfix in zip(self.projections,DIRS):
proj_path=path+"/"+postfix
utils.make_dir(proj_path)
full_path=proj_path+name
print(full_path)
utils.save_img(full_path,proj)
def final_to_pca(in_path, out_path):
# action=td.read_im_action(in_path+"/")
action = utils.read_object(in_path)
out_path = out_path.replace(".final", ".pca")
pca = action.to_pca()
eigen = td.to_time_serie(pca)
img = td.to_img(eigen)
utils.save_img(out_path, img)
def main():
check_version()
parser = build_parser()
options = parser.parse_args()
check_opts(options)
style_target = get_img(options.style)
if not options.slow:
content_targets = _get_files(options.train_path)
elif options.test:
content_targets = [options.test]
kwargs = {
"slow":options.slow,
"epochs":options.epochs,
"print_iterations":options.checkpoint_iterations,
"batch_size":options.batch_size,
"save_path":os.path.join(options.checkpoint_dir,'fns.ckpt'),
"learning_rate":options.learning_rate,
"device":options.device,
"total_iterations":options.total_iterations,
"base_model_path":options.base_model_path,
}
if options.slow:
if options.epochs < 10:
kwargs['epochs'] = 1000
if options.learning_rate < 1:
kwargs['learning_rate'] = 1e1
args = [
content_targets,
style_target,
options.content_weight,
options.style_weight,
options.tv_weight,
options.vgg_path
]
for preds, losses, i, epoch in optimize(*args, **kwargs):
style_loss, content_loss, tv_loss, loss = losses
print('Epoch %d, Iteration: %d, Loss: %s' % (epoch, i, loss))
to_print = (style_loss, content_loss, tv_loss)
print('style: %s, content:%s, tv: %s' % to_print)
sys.stdout.flush()
if options.test:
assert options.test_dir != False
preds_path = '%s/%s_%s.png' % (options.test_dir,epoch,i)
if not options.slow:
ckpt_dir = os.path.dirname(options.checkpoint_dir)
evaluate.ffwd_to_img(options.test,preds_path,
options.checkpoint_dir)
else:
save_img(preds_path, img)
ckpt_dir = options.checkpoint_dir
cmd_text = 'python evaluate.py --checkpoint-dir %s ...' % ckpt_dir
print("Training complete. For evaluation:\n `%s`" % cmd_text)
def main():
parser = build_parser()
options = parser.parse_args()
check_opts(options)
style_target = get_img(options.style)
if not options.slow:
content_targets = _get_files(options.train_path)
elif options.test:
content_targets = [options.test]
kwargs = {
"slow": options.slow,
"epochs": options.epochs,
"print_iterations": options.checkpoint_iterations,
"batch_size": options.batch_size,
"save_path": os.path.join(options.checkpoint_dir, 'fns.ckpt'),
"learning_rate": options.learning_rate,
"device_and_number": options.device_and_number
}
if options.slow:
if options.epochs < 10:
kwargs['epochs'] = 1000
if options.learning_rate < 1:
kwargs['learning_rate'] = 1e1
args = [
content_targets, style_target, options.content_weight,
options.style_weight, options.tv_weight, options.vgg_path
]
import time
from datetime import datetime
start_time = time.time()
for preds, losses, i, epoch in optimize(*args, **kwargs):
style_loss, content_loss, tv_loss, loss = losses
delta_time, start_time = time.time() - start_time, time.time()
print(
'Current Time = {}; Time Elapsed = {}; Epoch = {}; Iteration = {}; Loss = {}'
.format(datetime.now().strftime("%Y %B %d, %H:%M:%S"), delta_time,
epoch, i, loss))
to_print = (style_loss, content_loss, tv_loss)
print('Loss values: style = %s; content = %s; tv = %s' % to_print)
sys.stdout.flush()
if options.test:
assert options.test_dir != False
preds_path = '%s/%s_%s.png' % (options.test_dir, epoch, i)
if not options.slow: # if uses GPU, uses RAM that it doesn't have, so it's slow here
ckpt_dir = os.path.dirname(options.checkpoint_dir)
evaluate.ffwd_to_img(options.test, preds_path,
options.checkpoint_dir)
else:
save_img(preds_path, img)
ckpt_dir = options.checkpoint_dir
cmd_text = 'python evaluate.py --checkpoint %s ...' % ckpt_dir
print("Training complete. For evaluation:\n `%s`" % cmd_text)
def main():
print('ml5.js Style Transfer Training!')
print('Note: This traning will take a couple of hours.')
parser = build_parser()
options = parser.parse_args()
check_opts(options)
style_target = get_img(options.style)
if not options.slow:
content_targets = _get_files(options.train_path)
elif options.test:
content_targets = [options.test]
kwargs = {
"slow": options.slow,
"epochs": options.epochs,
"print_iterations": options.checkpoint_iterations,
"batch_size": options.batch_size,
"save_path": os.path.join(options.checkpoint_dir, 'fns.ckpt'),
"learning_rate": options.learning_rate,
}
if options.slow:
if options.epochs < 10:
kwargs['epochs'] = 1000
if options.learning_rate < 1:
kwargs['learning_rate'] = 1e1
args = [
content_targets, style_target, options.content_weight,
options.style_weight, options.tv_weight, options.vgg_path
]
print('Training is starting!...')
for preds, losses, i, epoch in optimize(*args, **kwargs):
style_loss, content_loss, tv_loss, loss = losses
print('Epoch %d, Iteration: %d, Loss: %s' % (epoch, i, loss))
to_print = (style_loss, content_loss, tv_loss)
print('style: %s, content:%s, tv: %s' % to_print)
if options.test:
assert options.test_dir != False
preds_path = '%s/%s_%s.png' % (options.test_dir, epoch, i)
if not options.slow:
ckpt_dir = os.path.dirname(options.checkpoint_dir)
evaluate.ffwd_to_img(options.test, preds_path,
options.checkpoint_dir)
else:
save_img(preds_path, img)
ckpt_dir = options.checkpoint_dir
cmd_text = 'python evaluate.py --checkpoint %s ...' % ckpt_dir
print("Training complete. For evaluation:\n `%s`" % cmd_text)
print('Converting model to ml5js')
dump_checkpoints(kwargs['save_path'], options.model_dir)
print(
'Done! Checkpoint saved. Visit https://ml5js.org/docs/StyleTransfer for more information'
)
def main():
parser = build_parser()
options = parser.parse_args()
check_opts(options)
content_img = get_img(options.content, (256, 256, 3)).astype(np.float32)
content_img = np.reshape(content_img, (1, ) + content_img.shape)
prediction = ffwd(content_img, options.style)
save_img(options.output_path, prediction)
print('Image saved to {}'.format(options.output_path))
def ffwd(data_in, paths_out, checkpoint_dir, device_t='/gpu:0', batch_size=4):
assert len(paths_out) > 0
is_paths = type(data_in[0]) == str
if is_paths:
assert len(data_in) == len(paths_out)
img_shape = get_img(data_in[0]).shape
else:
assert data_in.size[0] == len(paths_out)
img_shape = X[0].shape
g = tf.Graph()
batch_size = min(len(paths_out), batch_size)
curr_num = 0
soft_config = tf.ConfigProto(allow_soft_placement=True)
soft_config.gpu_options.allow_growth = True
with g.as_default(), g.device(device_t), \
tf.Session(config=soft_config) as sess:
batch_shape = (batch_size,) + img_shape
img_placeholder = tf.placeholder(tf.float32, shape=batch_shape,
name='img_placeholder')
preds = transform.net(img_placeholder)
saver = tf.train.Saver()
if os.path.isdir(checkpoint_dir):
ckpt = tf.train.get_checkpoint_state(checkpoint_dir)
if ckpt and ckpt.model_checkpoint_path:
saver.restore(sess, ckpt.model_checkpoint_path)
else:
raise Exception("No checkpoint found...")
else:
saver.restore(sess, checkpoint_dir)
num_iters = int(len(paths_out)/batch_size)
for i in range(num_iters):
pos = i * batch_size
curr_batch_out = paths_out[pos:pos+batch_size]
if is_paths:
curr_batch_in = data_in[pos:pos+batch_size]
X = np.zeros(batch_shape, dtype=np.float32)
for j, path_in in enumerate(curr_batch_in):
img = get_img(path_in)
assert img.shape == img_shape, \
'Images have different dimensions. ' + \
'Resize images or use --allow-different-dimensions.'
X[j] = img
else:
X = data_in[pos:pos+batch_size]
_preds = sess.run(preds, feed_dict={img_placeholder:X})
for j, path_out in enumerate(curr_batch_out):
save_img(path_out, _preds[j])
remaining_in = data_in[num_iters*batch_size:]
remaining_out = paths_out[num_iters*batch_size:]
if len(remaining_in) > 0:
ffwd(remaining_in, remaining_out, checkpoint_dir,
device_t=device_t, batch_size=1)
def ffwd(data_in, paths_out, checkpoint_dir, device_t='/gpu:0', batch_size=4):
assert len(paths_out) > 0
is_paths = type(data_in[0]) == str
if is_paths:
assert len(data_in) == len(paths_out)
img_shape = get_img(data_in[0]).shape
else:
assert data_in.size[0] == len(paths_out)
img_shape = X[0].shape
g = tf.Graph()
batch_size = min(len(paths_out), batch_size)
curr_num = 0
soft_config = tf.ConfigProto(allow_soft_placement=True)
soft_config.gpu_options.allow_growth = True
with g.as_default(), g.device(device_t), \
tf.Session(config=soft_config) as sess:
batch_shape = (batch_size,) + img_shape
img_placeholder = tf.placeholder(tf.float32, shape=batch_shape,
name='img_placeholder')
preds = transform.net(img_placeholder)
saver = tf.train.Saver()
if os.path.isdir(checkpoint_dir):
ckpt = tf.train.get_checkpoint_state(checkpoint_dir)
if ckpt and ckpt.model_checkpoint_path:
saver.restore(sess, ckpt.model_checkpoint_path)
else:
raise Exception("No checkpoint found...")
else:
saver.restore(sess, checkpoint_dir)
num_iters = int(len(paths_out)/batch_size)
for i in range(num_iters):
pos = i * batch_size
curr_batch_out = paths_out[pos:pos+batch_size]
if is_paths:
curr_batch_in = data_in[pos:pos+batch_size]
X = np.zeros(batch_shape, dtype=np.float32)
for j, path_in in enumerate(curr_batch_in):
img = get_img(path_in)
assert img.shape == img_shape, \
'Images have different dimensions. ' + \
'Resize images or use --allow-different-dimensions.'
X[j] = img
else:
X = data_in[pos:pos+batch_size]
_preds = sess.run(preds, feed_dict={img_placeholder:X})
for j, path_out in enumerate(curr_batch_out):
save_img(path_out, _preds[j])
remaining_in = data_in[num_iters*batch_size:]
remaining_out = paths_out[num_iters*batch_size:]
if len(remaining_in) > 0:
ffwd(remaining_in, remaining_out, checkpoint_dir,
device_t=device_t, batch_size=1)
def train(self):
for n in range(FLAGS.epochs):
for _ in tqdm(
iterable=range(FLAGS.steps_per_epoch),
ncols=int(get_terminal_width() * .9),
desc=tqdm.write(f'Epoch {n + 1}/{FLAGS.epochs}'),
unit=' steps',
):
self.train_step(self.img)
save_img(self.img.read_value(), n + 1)
def main():
parser = build_parser()
options = parser.parse_args()
if not os.path.exists(options.test_dir):
os.mkdir(options.test_dir)
if not os.path.exists(options.checkpoint_dir):
os.mkdir(options.checkpoint_dir)
check_opts(options)
style_target = get_img(options.style)
if not options.slow:
content_targets = _get_files(options.train_path)
elif options.test:
content_targets = [options.test]
kwargs = {
"slow": options.slow,
"epochs": options.epochs,
"print_iterations": options.checkpoint_iterations,
"batch_size": options.batch_size,
"save_path": os.path.join(options.checkpoint_dir, 'fns.ckpt'),
"learning_rate": options.learning_rate,
"gpu_fraction": options.gpu_fraction
}
if options.slow:
if options.epochs < 10:
kwargs['epochs'] = 1000
if options.learning_rate < 1:
kwargs['learning_rate'] = 1e1
args = [
content_targets, style_target, options.content_weight,
options.style_weight, options.tv_weight, options.vgg_path
]
for preds, losses, i, epoch in optimize(*args, **kwargs):
style_loss, content_loss, tv_loss, loss = losses
print('Epoch %d, Iteration: %d, Loss: %s' % (epoch, i, loss))
to_print = (style_loss, content_loss, tv_loss)
print('style: %s, content:%s, tv: %s' % to_print)
if options.test:
assert options.test_dir != False
preds_path = '%s/%s_%s.png' % (options.test_dir, epoch, i)
if not options.slow:
ckpt_dir = os.path.dirname(options.checkpoint_dir)
evaluate.ffwd_to_img(options.test, preds_path,
options.checkpoint_dir)
else:
save_img(preds_path, img)
ckpt_dir = options.checkpoint_dir
cmd_text = 'python evaluate.py --checkpoint %s ...' % ckpt_dir
print("Training complete. For evaluation:\n `%s`" % cmd_text)
def _execute(self):
if not os.path.exists(self.path + 'field_evaluation/'):
os.makedirs(self.path + 'field_evaluation/')
print('EVALUATION OF POINT CLOUD STARTED')
#Initialize CloudFeaturizer class
cfP = cloud_featurizer_Parameters()
featurizer = CloudFeaturizer(self.dws_point_cloud, cfP.METRIC,
cfP.DIMENSIONS)
print('POINT FEATURIZATION')
#Compute features
features = featurizer.collect_cloud_features()
fkeys = list(features.keys())
print('POINT CLASSIFICATION')
#Initialize VegetationClassifier class
classifier = VegetationClassifier(self.dws_point_cloud, features)
#Compute classification
crop_label, valid_idx = classifier.classify_vegetation()
#sample valid (classified) points for further analysis
classified_points = self.dws_point_cloud[valid_idx, :]
#Visualize na save classification result
save_img(np.hstack((classified_points, crop_label.reshape(-1, 1))),
'classification', self.path + 'field_evaluation/', 90, 0)
print('EDGE DETECTION')
#Initialize EdgeDetector class
edP = edge_detector_Parameters()
detector = EdgeDetector(classified_points, crop_label, edP.METRIC,
edP.ENTROPY_QUANTILE, edP.K)
edge_label = detector.get_edge_points()
save_img(np.hstack((classified_points, edge_label.reshape(-1, 1))),
'edge', self.path + 'field_evaluation/', 90, 0)
valid_points = classified_points
valid_points[:, 0] = valid_points[:, 0] + self.header.min[0]
valid_points[:, 1] = valid_points[:, 1] + self.header.min[1]
valid_points[:, 2] = valid_points[:, 2] + self.header.min[2]
np.savetxt(
self.path + 'field_evaluation/' +
'{}_features.csv'.format(fkeys[0]), features[fkeys[0]])
np.savetxt(
self.path + 'field_evaluation/' +
'{}_features.csv'.format(fkeys[1]), features[fkeys[1]])
np.savetxt(self.path + 'field_evaluation/' + 'labels.csv',
np.hstack((valid_points, crop_label.reshape(-1, 1),
edge_label.reshape(-1, 1))),
delimiter=",")
print('EVALUATION OF POINT CLOUD FINISHED')
def work(in_files, out_files, device_id, total_device, device_idx):
"""
Stylized the images
This function supports multi-GPU transformation
Arg: in_files - The path list of input images
out_files - The path list of output images
device_id - The name of device which is following the rule that tensorflow makes
total_device - The total number of devices
devices_idx - The index of the current device
"""
global adopt_revision
with tf.Graph().as_default():
tf_config = tf.ConfigProto(allow_soft_placement=True)
tf_config.gpu_options.allow_growth = True
# Construct graph
img_ph = tf.placeholder(tf.float32, shape=image_shape)
if adopt_revision == True:
logits = SmallAutoEncoder(img_ph)
else:
logits = AutoEncoder(img_ph)
# Run
with tf.Session(config=tf_config) as sess:
with tf.device(device_id): # Adopt multi-GPU to transfer the image
sess.run(tf.global_variables_initializer())
saver = tf.train.Saver()
saver.restore(sess, model_path + model_name)
if total_device <= 1:
start = 0
end = int(len(in_files) // 1)
else:
start = device_idx * int(len(in_files) // total_device)
end = device_idx * int(
len(in_files) // total_device) + int(
len(in_files) // total_device)
conduct_time = time.time()
for i in range(start, end, 1):
# print("progress: ", i, ' / ', end - start, '\t proc: ', device_idx)
img_batch = np.ndarray(image_shape)
for j, img_path in enumerate(in_files[i:i + 1]):
img = get_img(img_path)
img_batch[j] = img
_style_result = sess.run(
[
logits,
], feed_dict={img_ph: img_batch / 255.0})
for j, img_path in enumerate(out_files[i:i + 1]):
save_img(img_path, _style_result[0][j])
conduct_time = time.time() - conduct_time
print("Conduct time: ", conduct_time)
def main_worker(opt):
# opt.device = torch.device('cuda')
model = RACNN(num_classes=opt.num_classes)
model = model.to(opt.device)
# model = torch.nn.DataParallel(model,device_ids=[0,1])
print(model)
cls_params = list(model.b1.parameters()) + list(
model.b2.parameters()) + list(model.classifier1.parameters()) + list(
model.classifier2.parameters())
apn_params = list(model.apn.parameters())
# optimizer = model.parameters()
criterion = CrossEntropyLoss().to(opt.device)
(train_loader, train_logger, optimizer_cls, optimizer_apn,
scheduler) = get_train_utils(opt, cls_params, apn_params)
val_loader, val_logger = get_val_utils(opt)
test_sample, _ = next(iter(val_loader))
tb_writer = SummaryWriter(log_dir=opt.result_path)
pretrainAPN(train_loader, optimizer_apn, opt, model, tb_writer)
# model = model.to(opt.device)
for i in range(opt.begin_epoch, opt.n_epochs + 1):
cls_train_epoch(i, train_loader, model, criterion, optimizer_cls,
opt.device, train_logger, tb_writer)
apn_train_epoch(i, train_loader, model, optimizer_apn, opt.device,
tb_writer)
if i % opt.checkpoint == 0:
save_file_path = opt.result_path / 'save_{}.pth'.format(i)
save_checkpoint(save_file_path, i, model, optimizer_cls,
optimizer_apn, scheduler)
# if i % 5 == 0:
prev_val_loss = val_epoch(i, val_loader, model, criterion, opt.device,
val_logger, tb_writer)
if opt.lr_scheduler == 'multistep':
scheduler.step()
elif opt.lr_scheduler == 'plateau':
scheduler.step(prev_val_loss)
test_sample = test_sample.to(opt.device)
_, _, _, crops = model(test_sample)
img = crops[0].data
# pic_path = str(opt.result_path)+'/samples/'
save_img(
img,
path='/home/zhaoliu/car_brand/racnn/results/samples/iter_{}@2x.jpg'
.format(i),
annotation=f' 2xstep = {i}')
def ffwd(data_in, paths_out, model, device_t='', batch_size=1):
assert len(paths_out) > 0
is_paths = type(data_in[0]) == str
# TODO：如果 data_in 是保存输入图像的文件路径，即 is_paths 为 True，则读入第一张图像，由于 pb 模型的输入维度为 1 × 256 × 256 × 3, 因此需将输入图像的形状调整为 256 × 256，并传递给 img_shape；
# 如果 data_in 是已经读入图像并转化成数组形式的数据，即 is_paths 为 False，则直接获取图像的 shape 特征 img_shape
if is_paths:
# Get the shape when loading the first image.
img_shape = get_img(data_in[0], (256, 256, 3)).shape
else:
# Get the shape from data_in[0] when the images have been loaded.
img_shape = data_in[0].shape
g = tf.Graph()
config = tf.ConfigProto(allow_soft_placement=True,
inter_op_parallelism_threads=1,
intra_op_parallelism_threads=1)
with g.as_default():
with tf.gfile.FastGFile(model, 'rb') as f:
graph_def = tf.GraphDef()
graph_def.ParseFromString(f.read())
tf.import_graph_def(graph_def, name='')
with tf.Session(config=config) as sess:
sess.run(tf.global_variables_initializer())
input_tensor = sess.graph.get_tensor_by_name('X_content:0')
output_tensor = sess.graph.get_tensor_by_name('add_37:0')
batch_size = 1
# TODO：读入的输入图像的数据格式为 HWC，还需要将其转换成 NHWC
batch_shape = (batch_size, ) + img_shape
num_iters = int(len(paths_out) / batch_size)
for i in range(num_iters):
pos = i * batch_size
curr_batch_out = paths_out[pos:pos + batch_size]
# TODO：如果 data_in 是保存输入图像的文件路径，则依次将该批次中输入图像文件路径下的 batch_size 张图像读入数组 X；
# 如果 data_in 是已经读入图像并转化成数组形式的数据，则将该数组传递给 X
if is_paths:
curr_batch_in = data_in[pos:pos + batch_size]
X = np.zeros(batch_shape)
#print(X.shape)
for j, path in enumerate(curr_batch_in):
img = get_img(path, img_shape)
X[j] = img
else:
X = data_in[pos:pos + batch_size]
start = time.time()
# TODO: 使用 sess.run 来计算 output_tensor
_preds = sess.run(output_tensor, feed_dict={input_tensor: X})
end = time.time()
for j, path_out in enumerate(curr_batch_out):
#TODO：在该批次下调用 utils.py 中的 save_img() 函数对所有风格迁移后的图片进行存储
save_img(path_out, _preds[j])
delta_time = end - start
print("Inference (CPU) processing time: %s" % delta_time)
def evaluate_img(img_in, img_path, ckpt):
img_shape = (256, 256, 3)
batch_shape = (1, 256, 256, 3)
soft_config = tf.ConfigProto(allow_soft_placement=True)
soft_config.gpu_options.allow_growth = True
with tf.Graph().as_default(), tf.Session(config=soft_config) as sess:
# Declare placeholders we'll feed into the graph
X_inputs = tf.placeholder(
tf.float32, shape=batch_shape, name='X_inputs')
# Define output node
preds = transform.net(X_inputs) # (1, 720, 720, 3)
tf.identity(preds[0], name='output')
# For restore training checkpoints (important)
saver = tf.train.Saver()
saver.restore(sess, ckpt) # run
"""
saver = tf.train.Saver()
if os.path.isdir(FLAGS.checkpoint_dir):
ckpt = tf.train.get_checkpoint_state(FLAGS.checkpoint_dir)
if ckpt and ckpt.model_checkpoint_path:
saver.restore(sess, ckpt.model_checkpoint_path) # run
else:
raise Exception("No checkpoint found...")
else:
ckpt = saver.restore(sess, FLAGS.checkpoint_dir)
"""
X = np.zeros(batch_shape, dtype=np.float32) # feed
img = get_img(img_in, img_shape)
X[0] = img
_preds = sess.run(preds, feed_dict={X_inputs: X})
save_img(img_path, _preds[0])
# Write graph.
start_time = time.time()
tf.train.write_graph(
sess.graph.as_graph_def(),
FLAGS.model_dir,
FLAGS.model_name + '.pb',
as_text=False)
tf.train.write_graph(
sess.graph.as_graph_def(),
FLAGS.model_dir,
FLAGS.model_name + '.pb.txt',
as_text=True)
end_time = time.time()
delta_time = end_time - start_time
print('Save pb and pb.txt done!, time:', delta_time)
def main():
parser = build_parser()
options = parser.parse_args()
check_opts(options)
files = list_files(options.train_path)
print('hihi')
for x in files:
in_path = os.path.join(options.train_path, x)
out_path = os.path.join(options.result_path, x)
img = get_img(in_path, (256,256,3))
save_img(out_path, img)
def debug(self):
for job in self._memory.get_jobs():
for i, face in enumerate(job.get_faces()):
path = "".join([
self.config.final_debug_path, job.people_name,
job.obj_name,
str(i), '.jpg'
])
save_img(path, face)
for job in self._memory.get_jobs():
job.debug()
def generate(args, print_args=True):
if print_args:
print(args)
model = Model.load(args.model_file)
image = load_img(args.image_file)
_, height, width = image.size()
max_size = max(height, width)
img = generate_image(model, model.nz, height, width)
save_img(img, 'generate.png')
def _add_job_with_faces(self, job, faces):
job.add_faces(faces)
if self.config.debug:
for face in faces:
dict_key = "".join([job.people_name, job.obj_name])
path = "".join([
self.config.debug_path, job.people_name, job.obj_name,
str(self.find_obj_times[dict_key]), '.jpg'
])
save_img(path, face)
self.find_obj_times[dict_key] += 1
self._memory.add_job(job)
def main():
parser = build_parser()
options = parser.parse_args()
check_opts(options)
style_target = get_img(options.style)
if not options.slow:
content_targets = _get_files(options.train_path)
elif options.test:
content_targets = [options.test]
kwargs = {
"slow": options.slow,
"epochs": options.epochs,
"print_iterations": options.checkpoint_iterations,
"batch_size": options.batch_size,
"save_path": os.path.join(options.checkpoint_dir, 'fns.ckpt'),
"learning_rate": options.learning_rate
}
if options.slow:
if options.epochs < 10:
kwargs['epochs'] = 1000
if options.learning_rate < 1:
kwargs['learning_rate'] = 1e1
args = [
content_targets, style_target, options.content_weight,
options.style_weight, options.tv_weight, options.vgg_path
]
# with tf.Graph().as_default(), tf.device('/cpu:0'), tf.Session() as sess:
# init_op = tf.global_variables_initializer()
# saver = tf.train.Saver()
# saver.restore(sess, options.checkpoint_dir)
print('options.test: ' + options.test)
print('options.checkpoint_der: ' + options.checkpoint_dir)
for index_image in range(0, 20):
options.test = '../image_style_dataset/' + 'structure' + str(
index_image + 1).zfill(4) + '.jpg'
print(options.test)
preds_path = '%s/' % (options.test_dir) + SAVE_NAME + '%s_%s.png' % (
999, index_image)
if not options.slow:
ckpt_dir = os.path.dirname(options.checkpoint_dir)
evaluate.ffwd_to_img(options.test, preds_path,
options.checkpoint_dir)
else:
save_img(preds_path, img)
def on_epoch_end(self, epoch, logs=None):
if epoch % self.step != 0:
return
result = tf.squeeze(self.model(self.input_net), axis=0)
if self.rootdir is not None:
utils.save_img(
os.path.join(self.rootdir, f"epoch_{epoch}.png"),
result,
self.data_format,
)
if self.keep_output:
self.outputs_net.append(result)
if self.plot:
utils.plot_img(result, ncols=1, data_format=self.data_format)
def get_stylize_image(content_fullpath, style_fullpath, output_path,
content_size=256, style_size=256, alpha=0.6,
swap5=False, ss_alpha=0.6, adain=False):
content_img = get_img(content_fullpath)
content_img = resize_to(content_img, int(content_size))
style_img = get_img(style_fullpath)
style_img = resize_to(style_img,int(style_size))
stylized_rgb = wct_model.predict(
content_img, style_img, alpha, swap5, ss_alpha, adain)
save_img(output_path, stylized_rgb)
print("stylized image saved "+output_path)
def main():
parser = build_parser()
options = parser.parse_args()
check_opts(options)
# TODO：获取风格图像 style_target 以及内容图像数组 content_targets
style_target = ___________________
if not options.slow:
content_targets = _get_files(options.train_path)
elif options.test:
content_targets = [options.test]
kwargs = {
# "slow":options.slow,
"epochs": options.epochs,
"print_iterations": options.checkpoint_iterations,
"batch_size": options.batch_size,
"save_path": os.path.join(options.checkpoint_dir, 'fns.ckpt'),
"learning_rate": options.learning_rate,
"type": options.type,
"save": options.save
}
if options.slow:
if options.epochs < 10:
kwargs['epochs'] = 1000
if options.learning_rate < 1:
kwargs['learning_rate'] = 1e1
args = [
content_targets, style_target, options.content_weight,
options.style_weight, options.tv_weight, options.vgg_path
]
for preds, losses, i, epoch in optimize(*args, **kwargs):
style_loss, content_loss, tv_loss, loss = losses
print('Epoch %d, Iteration: %d, Loss: %s' % (epoch, i, loss))
to_print = (style_loss, content_loss, tv_loss)
print('style: %s, content:%s, tv: %s' % to_print)
if options.test:
assert options.test_dir != False
preds_path = '%s/%s_%s.png' % (options.test_dir, epoch, i)
if not options.slow:
ckpt_dir = os.path.dirname(options.checkpoint_dir)
evaluate.ffwd_to_img(options.test, preds_path,
options.checkpoint_dir)
else:
save_img(preds_path, img)
ckpt_dir = options.checkpoint_dir
print("Training complete.")
def test_single(self, img_fn):
# networks
self.model = Net(num_channels=self.num_channels,
base_filter=64,
num_residuals=18)
if self.gpu_mode:
self.model.cuda()
# load model
self.load_model()
# load data
img = Image.open(img_fn)
img = img.convert('YCbCr')
y, cb, cr = img.split()
y = y.resize(
(y.size[0] * self.scale_factor, y.size[1] * self.scale_factor),
Image.BICUBIC)
input = Variable(ToTensor()(y)).view(1, -1, y.size[1], y.size[0])
if self.gpu_mode:
input = input.cuda()
self.model.eval()
recon_img = self.model(input)
# save result images
utils.save_img(recon_img.cpu().data, 1, save_dir=self.save_dir)
out = recon_img.cpu()
out_img_y = out.item()
out_img_y = (((out_img_y - out_img_y.min()) * 255) /
(out_img_y.max() - out_img_y.min())).numpy()
# out_img_y *= 255.0
# out_img_y = out_img_y.clip(0, 255)
out_img_y = Image.fromarray(np.uint8(out_img_y[0]), mode='L')
out_img_cb = cb.resize(out_img_y.size, Image.BICUBIC)
out_img_cr = cr.resize(out_img_y.size, Image.BICUBIC)
out_img = Image.merge(
'YCbCr', [out_img_y, out_img_cb, out_img_cr]).convert('RGB')
# save img
result_dir = os.path.join(self.save_dir, 'result')
if not os.path.exists(result_dir):
os.mkdir(result_dir)
save_fn = result_dir + '/SR_result.png'
out_img.save(save_fn)
def forward_prop(data_in,
paths_out,
checkpoint_dir,
device_t='/cpu:0',
batch_size=1):
assert len(paths_out) > 0
is_paths = type(data_in) == str
img_shape = get_img(data_in).shape
#print("Batch size: ", batch_size)
g = tf.Graph()
soft_config = tf.ConfigProto(allow_soft_placement=True)
soft_config.gpu_options.allow_growth = True
with g.as_default(), g.device(device_t), tf.Session(
config=soft_config) as sess:
batch_shape = (batch_size, ) + img_shape
#print("Batch_shape: ", batch_shape)
img_placeholder = tf.placeholder(tf.float32,
shape=batch_shape,
name='img_placeholder')
preds = transform.net(img_placeholder)
saver = tf.train.Saver()
#Restore checkpoint in session
saver.restore(sess, checkpoint_dir)
curr_batch_out = paths_out
if is_paths:
curr_batch_in = data_in
print("curr_batch_in: ", curr_batch_in)
print("curr_batch_out: ", curr_batch_out)
X = np.zeros(batch_shape, dtype=np.float32)
img = get_img(curr_batch_in)
assert img.shape == img_shape, 'Images have different dimensions. ' + 'Resize images'
X[0] = img
#print("Shape: ", X.shape) #(1,960,960,3)
_preds = sess.run(preds, feed_dict={img_placeholder: X})
save_img(curr_batch_out, _preds[0])
sess.close()
print("Done!!")
return curr_batch_out
def create_latent_images():
model = load_model("model.h5")
data_dir = TRAIN_PATH
shutil.rmtree("result/")
dir_names = sorted(os.listdir(data_dir))
generator = data_gen(data_dir, 1)
for i in range(len(dir_names) - 1):
(encoder_input, decoder_input), _ = next(generator)
latent_image = model.predict([encoder_input, decoder_input])
target_dir = os.path.join("result/", dir_names[i + 1])
os.makedirs(target_dir)
for j in range(latent_image.shape[1]):
img = latent_image[0][j]
save_path = os.path.join(target_dir, "{0:05d}.jpg".format(j + 1))
save_img(img, save_path)
def apply_cls(dim,s_cat,params,actions,extr):
imgs=data.get_named_projections(dim,actions)
img_cats=[]
for img_i in imgs:
img_j=np.reshape(img_i[1],(1,img_i[1].size))
cat=extr.test(img_j)
if(s_cat==cat):
img_cats.append(img_i)
out_path=params['out_path']
out_path=out_path+str(s_cat)+"/"
utils.make_dir(out_path)
for cat,img_i in img_cats:
full_path=out_path+cat
img_i=np.reshape(img_i, (60,60))
utils.save_img(full_path,img_i)
def __init__(self, n_scans, in_dir, out_dir):
self.in_dir = in_dir
self.out_dir = out_dir
self.in_file_paths = glob(in_dir)
self.in_file_paths.sort()
self.rx_number = 1
self.rx_component = 'Ez'
self.n_scans = n_scans
for path in tqdm(self.in_file_paths):
self._create_ascan(path)
self._merge_ascan(path)
outputdata, dt = self._prepare_bscan(path)
save_img(path, outputdata, dt, self.rx_number, self.rx_component,
self.out_dir)
def build(self):
#mask initialization
b, c, w, h = self.original_img_tensor.shape
mask_tensor = torch.rand(
(b, 1, int(w / self.factor), int(h / self.factor)))
if cuda_available():
mask_tensor = mask_tensor.cuda()
mask_tensor = Variable(mask_tensor, requires_grad=True)
output = self.model(self.original_img_tensor)
#target class for explanations
class_index = np.argmax(output.data.cpu().numpy())
optimizer = torch.optim.Adam([mask_tensor], self.lr)
for i in range(self.iter + 1):
#upsampling mask to fit the shape of mask to the shape of image
upsampled_mask = self.upsample(mask_tensor)
#gjttering
jitter = torch.randn((b, c, w, h)) * 0.03
jitter_org_img_tensor = self.original_img_tensor + jitter.cuda()
mask_img=torch.mul(upsampled_mask,jitter_org_img_tensor)+torch.mul((1-upsampled_mask),\
self.perturbed_img_tensor)
mask_output = torch.nn.Softmax(dim=1)(self.model(mask_img))
mask_prob = mask_output[0, class_index]
loss=self.l1_coeff*torch.mean(1-torch.abs(mask_tensor))+\
TV(mask_tensor,self.tv_coeff,self.tv_beta)+mask_prob
optimizer.zero_grad()
loss.backward()
optimizer.step()
#allow the values of mask to be [0,1]
mask_tensor.data.clamp_(0, 1)
if i % 20 == 0:
print(
f'[{i}/{self.iter}] Loss: {loss} Prob for the target class: {mask_prob}'
)
save_img(self.upsample(mask_tensor), self.original_img,
self.perturbed_img, self.img_path, self.model_path)
def train(self, dataloader, epochs=100):
optim_D = optim.Adam(self.D.parameters(), lr=1e-4)
optim_G = optim.Adam(self.G.parameters(), lr=1e-4)
for epoch in range(epochs):
bar = tqdm(dataloader)
for batch, (data, labels) in enumerate(bar):
### train_D
self.D.train()
self.G.eval()
optim_D.zero_grad()
x_real = data.to(self.device)
x_fake = self.generate_fake(self.batch_size)[0]
predict_real = self.D(x_real)
predict_fake = self.D(x_fake)
loss_D_real = -torch.log(predict_real).mean()
loss_D_fake = -torch.log(torch.clamp(1 - predict_fake,
min=eps)).mean()
loss_D = loss_D_real + loss_D_fake
loss_D.backward()
optim_D.step()
acc_D = (torch.sum(predict_real > 0.5).item() + torch.sum(
predict_fake < 0.5).item()) / self.batch_size / 2
### train_G
self.G.train()
self.D.eval()
optim_G.zero_grad()
x_fake = self.generate_fake(self.batch_size)[0]
predict_fake = self.D(x_fake)
loss_G = -torch.log(predict_fake).mean()
loss_G.backward()
optim_G.step()
acc_G = torch.sum(
predict_fake < 0.5).item() / predict_fake.shape[0]
bar.set_postfix(loss_D='{:.03f}'.format(loss_D),
acc_D='{:.03f}'.format(acc_D),
loss_G='{:.03f}'.format(loss_G),
acc_G='{:.03f}'.format(acc_G))
img = self.generate_fake(1)[0][0].detach().cpu()
save_img(img, f'output/{epoch}.jpg')
def storeTensor(tensor, store_path = "?"):
"""
This function assume the size of tensor is [32, x, x, 1]
"""
result_img = None
for i in range(8):
row = None
for j in range(4):
if row is None:
row = tensor[i*8+j]
else:
row = np.concatenate((row, tensor[i*8+j]), axis=2)
if result_img is None:
result_img = row
else:
result_img = np.concatenate((result_img, row), axis=1)
save_img(store_path, result_img)
def overlay(img: Image, overlay_color: tuple, orig_file_name:str):
"""
Place an overlay over an existing image
:param img: Image opened with PIL.Image
:param overlay_color: four-tuple with color to add to your image
:param orig_file_name: name of the original file
"""
assert len(overlay_color) == 4, 'Overlay color shall be a 4-tuple'
img_overlay = Image.new(size=img.size, color=overlay_color, mode='RGBA')
img.paste(img_overlay, None, mask=img_overlay)
color_string = '_'.join([str(c) for c in overlay_color])
filename = '{orig}_overlay_{color}.jpg'.format(orig=orig_file_name, color=color_string)
save_img(img, filename)
def save_neg_samples(root_folder, folder_to, train=True, subset_size=None, generate_subset=False):
"""Se encarga de cargar todos los samples negativos"""
lst_file_folder = 'Train' if train else 'Test'
lst_neg_file = os.path.join(lst_file_folder, 'neg.lst')
content_neg = open(os.path.join(root_folder, lst_neg_file)) # Abro el listado de imagenes positivas
content_neg_lines = content_neg.readlines()
if subset_size:
random.shuffle(content_neg_lines) # Los pongo en orden aleatorio cuando genero subset
content_neg_lines = content_neg_lines[0:subset_size] # Si fue especificado un tamaño de subset recorto el dataset
for img_path in content_neg_lines:
img_path = img_path.rstrip('\n') # Cuando lee la linea queda el \n en el final, lo eliminamos
img = skimage.io.imread(os.path.join(root_folder, img_path)) # Cargo la imagen
if generate_subset and not subset_size:
utils.generate_sub_samples(img, img_path, folder_to) # Genero nuevas muestras a partir de la imagen
img = utils.resize(img) # Re escalo la imagen original
filename = utils.get_filename(img_path) # Genero el nombre que tendra la imagen guardada
utils.save_img(img, folder_to, filename) # Guardo la imagen en la carpeta de negativos
def test(self):
print("Begin test...")
# load trained model
self.model.load(self.sess)
# inference validation images & calculate PSNR
mean_runtime= 0
mean_psnr = 0
file_name_list = self.data.file_names_for_dirs[0]
for (in_img, file_name) in zip(self.data.dataset[0], file_name_list):
# inference
start_time = time.time()
if self.args.degrade or self.args.no_self_ensemble:
out_img = chop_forward(in_img, self.sess, self.model, scale=self.data.scale_list[0], shave=10)
# Only for Track 1, we use geometric self-ensemble
else:
tmp_img = np.zeros([in_img.shape[0]*self.data.scale_list[0], in_img.shape[1]*self.data.scale_list[0], 3])
for i in range(2):
if i == 0:
flip_img = np.fliplr(in_img)
for j in range(4):
rot_flip_img = np.rot90(flip_img, j)
out_img = chop_forward(rot_flip_img, self.sess, self.model, scale=self.data.scale_list[0], shave=10)
tmp_img += np.fliplr(np.rot90(out_img, 4-j))
else:
for k in range(4):
rot_img = np.rot90(in_img, k)
out_img = chop_forward(rot_img, self.sess, self.model, scale=self.data.scale_list[0], shave=10)
tmp_img += np.rot90(out_img, 4-k)
out_img = tmp_img / 8
end_time = time.time()
mean_runtime += (end_time - start_time) / self.args.num_test
# save images
dir_img = os.path.join(self.args.exp_dir, self.args.exp_name, 'results', file_name)
save_img(out_img, dir=dir_img)
# write text file for summarize results
dir_file = os.path.join(self.args.exp_dir, self.args.exp_name, 'results', 'results.txt')
with open(dir_file, "w") as f:
f.write("runtime per image [s] : {0:2.2f}\n".format(mean_runtime))
f.write("CPU[1] / GPU[0] : 0\n")
f.write("Extra Data [1] / No Extra Data [0] : 0")
print("Test is done!")
def test(data_loader, model, logger, epoch):
with torch.no_grad():
metric = Metric()
model.train(False)
for i, input in enumerate(data_loader):
input = collate(input)
input_size = input['img'].size(0)
input = to_device(input, cfg['device'])
output = model(input)
output['loss'] = output['loss'].mean() if cfg['world_size'] > 1 else output['loss']
evaluation = metric.evaluate(cfg['metric_name']['test'], input, output)
logger.append(evaluation, 'test', input_size)
logger.append(evaluation, 'test')
info = {'info': ['Model: {}'.format(cfg['model_tag']), 'Test Epoch: {}({:.0f}%)'.format(epoch, 100.)]}
logger.append(info, 'test', mean=False)
logger.write('test', cfg['metric_name']['test'])
save_img(input['img'][:100], './output/vis/input_{}.png'.format(cfg['model_tag']), range=(-1, 1))
save_img(output['img'][:100], './output/vis/output_{}.png'.format(cfg['model_tag']), range=(-1, 1))
return
def save_neg_samples(folder_from, folder_to, subset_size=None):
"""Se encarga de cargar todos los samples negativos"""
for dirpath, dirnames, filenames in os.walk(
folder_from): # Obtengo los nombres de los archivos
if subset_size:
random.shuffle(
filenames) # Los pongo en orden aleatorio cuando genero subset
filenames = filenames[
0:
subset_size] # Si fue especificado un tamaño de subset recorto el dataset
for filename in filenames:
img_path = os.path.join(dirpath, filename)
img = skimage.io.imread(img_path) # Cargo la imagen
# generate_sub_samples(img, img_path) # Genero nuevas muestras a partir de la imagen
img = utils.resize(img) # Re escalo la imagen original
basename, extension = utils.get_basename(filename)
filename_final = basename + '.png'
utils.save_img(
img, folder_to,
filename_final) # Guardo la imagen en la carpeta de negativos
for y in range(height):
# Get new width at this height
relative_y = (y / height - 0.5) * 2 # Convert to value between -1 and +1
angle = math.asin(relative_y)
relative_x = math.cos(angle)
new_width = int(math.ceil(relative_x * width))
# Cut a strip at this height
left, right = 0, width
upper, lower = y, y + 1
box = (left, upper, right, lower)
img_strip = img.crop(box)
# Resize the strip
img_resized = img_strip.resize(size=(new_width, 1))
# Paste the resized strip in place
left = int(round(0.5 * (width - new_width)))
upper = y
box = (left, upper)
result.paste(img_resized, box)
return result
if __name__ == '__main__':
orig_name = 'amsterdam_190x150'
ams = open_img('{}.jpg'.format(orig_name))
ams_circular = make_circular(ams)
save_img(ams_circular, '{}_circular.png'.format(orig_name))
"""
Created on Mar 17, 2018
Used to answer this question:
https://stackoverflow.com/questions/49263496/pil-image-rotate-center0-0
@author: physicalattraction
"""
from utils import open_img, save_img
if __name__ == '__main__':
orig_file_name = 'amsterdam_190x150'
img = open_img('{}.jpg'.format(orig_file_name))
angle = 45
# Note: the rotated image has black pixels at the edges
rotated_img = img.rotate(angle, expand=True)
save_img(rotated_img, 'rotate_{}_{}.jpg'.format(angle, orig_file_name))
def main():
# Load the WCT model
wct_model = WCT(checkpoints=args.checkpoints,
relu_targets=args.relu_targets,
vgg_path=args.vgg_path,
device=args.device,
ss_patch_size=args.ss_patch_size,
ss_stride=args.ss_stride)
# Create needed dirs
in_dir = os.path.join(args.tmp_dir, 'input')
out_dir = os.path.join(args.tmp_dir, 'sytlized')
if not os.path.exists(in_dir):
os.makedirs(in_dir)
if not os.path.exists(out_dir):
os.makedirs(out_dir)
if os.path.isdir(args.in_path):
in_path = get_files(args.in_path)
else: # Single image file
in_path = [args.in_path]
if os.path.isdir(args.style_path):
style_files = get_files(args.style_path)
else: # Single image file
style_files = [args.style_path]
print(style_files)
import time
# time.sleep(999)
in_args = [
'ffmpeg',
'-i', args.in_path,
'%s/frame_%%d.png' % in_dir
]
subprocess.call(" ".join(in_args), shell=True)
base_names = os.listdir(in_dir)
in_files = [os.path.join(in_dir, x) for x in base_names]
out_files = [os.path.join(out_dir, x) for x in base_names]
s = time.time()
for content_fullpath in in_path:
content_prefix, content_ext = os.path.splitext(content_fullpath)
content_prefix = os.path.basename(content_prefix)
try:
for style_fullpath in style_files:
style_img = get_img(style_fullpath)
if args.style_size > 0:
style_img = resize_to(style_img, args.style_size)
if args.crop_size > 0:
style_img = center_crop(style_img, args.crop_size)
style_prefix, _ = os.path.splitext(style_fullpath)
style_prefix = os.path.basename(style_prefix)
# print("ARRAY: ", style_img)
out_v = os.path.join(args.out_path, '{}_{}{}'.format(content_prefix, style_prefix, content_ext))
print("OUT:",out_v)
if os.path.isfile(out_v):
print("SKIP" , out_v)
continue
for in_f, out_f in zip(in_files, out_files):
print('{} -> {}'.format(in_f, out_f))
content_img = get_img(in_f)
if args.keep_colors:
style_rgb = preserve_colors_np(style_img, content_img)
else:
style_rgb = style_img
stylized = wct_model.predict(content_img, style_rgb, args.alpha, args.swap5, args.ss_alpha)
if args.passes > 1:
for _ in range(args.passes-1):
stylized = wct_model.predict(stylized, style_rgb, args.alpha)
# Stitch the style + stylized output together, but only if there's one style image
if args.concat:
# Resize style img to same height as frame
style_img_resized = scipy.misc.imresize(style_rgb, (stylized.shape[0], stylized.shape[0]))
stylized = np.hstack([style_img_resized, stylized])
save_img(out_f, stylized)
fr = 30
out_args = [
'ffmpeg',
'-i', '%s/frame_%%d.png' % out_dir,
'-f', 'mp4',
'-q:v', '0',
'-vcodec', 'mpeg4',
'-r', str(fr),
'"' + out_v + '"'
]
print(out_args)
subprocess.call(" ".join(out_args), shell=True)
print('Video at: %s' % out_v)
if args.keep_tmp is True or len(style_files) > 1:
continue
else:
shutil.rmtree(args.tmp_dir)
print('Processed in:',(time.time() - s))
print('Processed in:',(time.time() - s))
except Exception as e:
print("EXCEPTION: ",e)
Inputs:
-------
img: Image opened with PIL.Image
xy: tuple with relative (x,y) position of the center of the cropped image
x and y shall be between 0 and 1
scale_factor: the ratio between the original image's size and the cropped image's size
"""
center = (img.size[0] * xy[0], img.size[1] * xy[1])
new_size = (img.size[0] / scale_factor, img.size[1] / scale_factor)
left = max(0, int(center[0] - new_size[0] / 2))
right = min(img.size[0], int(center[0] + new_size[0] / 2))
upper = max(0, int(center[1] - new_size[1] / 2))
lower = min(img.size[1], int(center[1] + new_size[1] / 2))
cropped_img = img.crop((left, upper, right, lower))
return cropped_img
if __name__ == '__main__':
orig_file_name = 'amsterdam_190x150'
ams = open_img('{}.jpg'.format(orig_file_name))
crop_ams = crop_image(ams, (0.50, 0.50), 1.1)
save_img(crop_ams, 'crop_{}_01.jpg'.format(orig_file_name))
crop_ams = crop_image(ams, (0.25, 0.25), 2.5)
save_img(crop_ams, 'crop_{}_02.jpg'.format(orig_file_name))
crop_ams = crop_image(ams, (0.75, 0.45), 3.5)
save_img(crop_ams, 'crop_{}_03.jpg'.format(orig_file_name))
def main():
start = time.time()
# Load the WCT model
wct_model = WCT(checkpoints=args.checkpoints,
relu_targets=args.relu_targets,
vgg_path=args.vgg_path,
device=args.device,
ss_patch_size=args.ss_patch_size,
ss_stride=args.ss_stride)
# Get content & style full paths
if os.path.isdir(args.content_path):
content_files = get_files(args.content_path)
else: # Single image file
content_files = [args.content_path]
if os.path.isdir(args.style_path):
style_files = get_files(args.style_path)
if args.random > 0:
style_files = np.random.choice(style_files, args.random)
else: # Single image file
style_files = [args.style_path]
os.makedirs(args.out_path, exist_ok=True)
count = 0
### Apply each style to each content image
for content_fullpath in content_files:
content_prefix, content_ext = os.path.splitext(content_fullpath)
content_prefix = os.path.basename(content_prefix) # Extract filename prefix without ext
content_img = get_img(content_fullpath)
if args.content_size > 0:
content_img = resize_to(content_img, args.content_size)
for style_fullpath in style_files:
style_prefix, _ = os.path.splitext(style_fullpath)
style_prefix = os.path.basename(style_prefix) # Extract filename prefix without ext
# style_img = get_img_crop(style_fullpath, resize=args.style_size, crop=args.crop_size)
# style_img = resize_to(get_img(style_fullpath), content_img.shape[0])
style_img = get_img(style_fullpath)
if args.style_size > 0:
style_img = resize_to(style_img, args.style_size)
if args.crop_size > 0:
style_img = center_crop(style_img, args.crop_size)
if args.keep_colors:
style_img = preserve_colors_np(style_img, content_img)
# if args.noise: # Generate textures from noise instead of images
# frame_resize = np.random.randint(0, 256, frame_resize.shape, np.uint8)
# frame_resize = gaussian_filter(frame_resize, sigma=0.5)
# Run the frame through the style network
stylized_rgb = wct_model.predict(content_img, style_img, args.alpha, args.swap5, args.ss_alpha, args.adain)
if args.passes > 1:
for _ in range(args.passes-1):
stylized_rgb = wct_model.predict(stylized_rgb, style_img, args.alpha, args.swap5, args.ss_alpha, args.adain)
# Stitch the style + stylized output together, but only if there's one style image
if args.concat:
# Resize style img to same height as frame
style_img_resized = scipy.misc.imresize(style_img, (stylized_rgb.shape[0], stylized_rgb.shape[0]))
# margin = np.ones((style_img_resized.shape[0], 10, 3)) * 255
stylized_rgb = np.hstack([style_img_resized, stylized_rgb])
# Format for out filename: {out_path}/{content_prefix}_{style_prefix}.{content_ext}
out_f = os.path.join(args.out_path, '{}_{}{}'.format(content_prefix, style_prefix, content_ext))
# out_f = f'{content_prefix}_{style_prefix}.{content_ext}'
save_img(out_f, stylized_rgb)
count += 1
print("{}: Wrote stylized output image to {}".format(count, out_f))
print("Finished stylizing {} outputs in {}s".format(count, time.time() - start))
"""
Created on ???
Used to answer this question:
???
@author: physicalattraction
"""
from PIL import Image
from utils import save_img
def create_img(width: int = 200, height: int = 200) -> Image:
img = Image.new('RGB', (width, height))
pixel_list = [(i % 256, i % 256, i % 256) for i in range(width * height)]
i_pixel = 0
for x in range(width):
for y in range(height):
img.putpixel((x, y), pixel_list[i_pixel])
i_pixel += 1
return img
if __name__ == '__main__':
droopy = create_img()
save_img(droopy, 'droopy.png')
@author: physicalattraction
"""
from PIL import Image
import math
from utils import open_img, save_img
def spotlight(img: Image, center: (int, int), radius: int) -> Image:
width, height = img.size
overlay_color = (0, 0, 0, 128)
img_overlay = Image.new(size=img.size, color=overlay_color, mode='RGBA')
for x in range(width):
for y in range(height):
dx = x - center[0]
dy = y - center[1]
distance = math.sqrt(dx * dx + dy * dy)
if distance < radius:
img_overlay.putpixel((x, y), (0, 0, 0, 0))
img.paste(img_overlay, None, mask=img_overlay)
return img
if __name__ == '__main__':
orig_file_name = 'amsterdam_1900x1500'
img = open_img('{}.jpg'.format(orig_file_name))
spotlight_img = spotlight(img, (475, 900), 400)
save_img(spotlight_img, 'spotlight_{}.jpg'.format(orig_file_name))
def save(self,in_path):
for i,img_i in enumerate(self.imgs):
full_path=in_path+"_"+str(i)
if(i==0):
img_i=np.reshape(img_i,(60,60))
utils.save_img(full_path,img_i)
def main():
# Load the WCT model
wct_model = WCT(checkpoints=args.checkpoints,
relu_targets=args.relu_targets,
vgg_path=args.vgg_path,
device=args.device,
ss_patch_size=args.ss_patch_size,
ss_stride=args.ss_stride)
# Load a panel to control style settings
style_window = StyleWindow(args.style_path,
args.style_size,
args.crop_size,
args.scale,
args.alpha,
args.swap5,
args.ss_alpha,
args.passes)
# Start the webcam stream
cap = WebcamVideoStream(args.video_source, args.width, args.height).start()
_, frame = cap.read()
# Grab a sample frame to calculate frame size
frame_resize = cv2.resize(frame, None, fx=args.scale, fy=args.scale)
img_shape = frame_resize.shape
# Setup video out writer
if args.video_out is not None:
fourcc = cv2.VideoWriter_fourcc(*'XVID')
if args.concat:
out_shape = (img_shape[1]+img_shape[0],img_shape[0]) # Make room for the style img
else:
out_shape = (img_shape[1],img_shape[0])
print('Video Out Shape:', out_shape)
video_writer = cv2.VideoWriter(args.video_out, fourcc, args.fps, out_shape)
fps = FPS().start() # Track FPS processing speed
# Toggles changed with kb shortcuts
keep_colors = args.keep_colors
swap_style = args.swap5
use_adain = args.adain
count = 0
while(True):
ret, frame = cap.read()
if ret is True:
frame_resize = cv2.resize(frame, None, fx=style_window.scale, fy=style_window.scale)
if args.noise: # Generate textures from noise instead of images
frame_resize = np.random.randint(0, 256, frame_resize.shape, np.uint8)
count += 1
print("Frame:",count,"Orig shape:",frame.shape,"New shape",frame_resize.shape)
content_rgb = cv2.cvtColor(frame_resize, cv2.COLOR_BGR2RGB) # OpenCV uses BGR, we need RGB
if args.random > 0 and count % args.random == 0:
style_window.set_style(random=True)
if keep_colors:
style_rgb = preserve_colors_np(style_window.style_rgb, content_rgb)
else:
style_rgb = style_window.style_rgb
# Run the frame through the style network
stylized_rgb = wct_model.predict(content_rgb, style_rgb, style_window.alpha, swap_style, style_window.ss_alpha, use_adain)
# Repeat stylization pipeline
if style_window.passes > 1:
for i in range(style_window.passes-1):
stylized_rgb = wct_model.predict(stylized_rgb, style_rgb, style_window.alpha, swap_style, style_window.ss_alpha, use_adain)
# Stitch the style + stylized output together
if args.concat:
# Resize style img to same height as frame
style_rgb_resized = cv2.resize(style_rgb, (stylized_rgb.shape[0], stylized_rgb.shape[0]))
stylized_rgb = np.hstack([style_rgb_resized, stylized_rgb])
stylized_bgr = cv2.cvtColor(stylized_rgb, cv2.COLOR_RGB2BGR)
if args.video_out is not None:
stylized_bgr = cv2.resize(stylized_bgr, out_shape) # Make sure frame matches video size
video_writer.write(stylized_bgr)
cv2.imshow('WCT Universal Style Transfer', stylized_bgr)
fps.update()
key = cv2.waitKey(10)
if key & 0xFF == ord('r'): # Load new random style
style_window.set_style(random=True)
elif key & 0xFF == ord('c'): # Toggle color preservation
keep_colors = not keep_colors
print('Switching to keep_colors:',keep_colors)
elif key & 0xFF == ord('s'): # Toggle style swap
swap_style = not swap_style
print('New value for flag swap_style:',swap_style)
elif key & 0xFF == ord('a'): # Toggle AdaIN
use_adain = not use_adain
print('New value for flag use_adain:',use_adain)
elif key & 0xFF == ord('w'): # Write stylized frame
out_f = "{}.png".format(time.time())
save_img(out_f, stylized_rgb)
print('Saved image to:',out_f)
elif key & 0xFF == ord('q'): # Quit gracefully
break
else:
break
fps.stop()
print('[INFO] elapsed time (total): {:.2f}'.format(fps.elapsed()))
print('[INFO] approx. FPS: {:.2f}'.format(fps.fps()))
cap.stop()
if args.video_out is not None:
video_writer.release()
cv2.destroyAllWindows()
"""
Created on Oct 13, 2017
Used to answer this question:
https://stackoverflow.com/a/46736330/1469465
@author: physicalattraction
"""
from PIL import ImageDraw
from utils import save_img, open_img, print_pil_version_info
if __name__ == '__main__':
image = open_img('star_transparent.png')
width, height = image.size
center = (int(0.5 * width), int(0.5 * height))
yellow = (255, 255, 0, 255)
ImageDraw.floodfill(image, xy=center, value=yellow)
save_img(image, 'star_yellow.png')
print_pil_version_info()