def unfreeze_model():
"""unfreeze model"""
output_graph_path = os.path.join(FLAGS.model, FLAGS.out_graph_name)
# Ouput path
model_p = FLAGS.model
model_p = model_p if not model_p.endswith("/") else model_p[:-1]
model_p = os.path.split(model_p)
output_path = os.path.join("output", model_p[len(model_p) - 1])
with tf.Graph().as_default():
output_graph_def = tf.GraphDef()
with open(output_graph_path, "rb") as f:
output_graph_def.ParseFromString(f.read())
_ = tf.import_graph_def(output_graph_def, name="")
content_images = reader.get_image(FLAGS.content_image,
FLAGS.image_size)
images = tf.pack([content_images])
with tf.Session() as sess:
input_node = sess.graph.get_tensor_by_name("input_node:0")
output_node = sess.graph.get_tensor_by_name("output_node:0")
im_images = sess.run(images)
output = sess.run(output_node, feed_dict={input_node: im_images})
out_path = os.path.join(output_path,
FLAGS.output + '-unfreeze.png')
print("Save result in: ", out_path)
print('------------------------------------')
print('Finished!')
misc.imsave(out_path, output[0])
def execute(self, data, batch_size):
sess = self.output['sess']
graph = self.output['graph']
for i in range(batch_size):
height = 0
width = 0
with open(data[i], 'rb') as img:
if data[i].lower().endswith('png'):
im = sess.run(tf.image.decode_png(img.read()))
else:
im = sess.run(tf.image.decode_jpeg(img.read()))
height = im.shape[0]
width = im.shape[0]
tf.logging.info('Image size: %dx%d' % (width, height))
image = reader.get_image(data[i], height, width,
self.output['image_preprocessing_fn'])
image = tf.expand_dims(image, 0)
generated = model.net(image, training=False)
generated = tf.cast(generated, tf.uint8)
generated = tf.squeeze(generated, [0])
# Save it to UCloud
generated_file = 'generated/res.jpg'
if os.path.exists('generated') is False:
os.makedirs('generated')
with open(generated_file, 'wb') as img:
start_time = time.time()
img.write(sess.run(tf.image.encode_jpeg(generated)))
end_time = time.time()
tf.logging.info('Elapsed time: %fs' % (end_time - start_time))
tf.logging.info('Done. Please check %s.' % generated_file)
def main(_):
# Get image's height and width.
height = 0
width = 0
with open(FLAGS.image_file, 'rb') as img:
with tf.Session().as_default() as sess:
if FLAGS.image_file.lower().endswith('png'):
image = sess.run(tf.image.decode_png(img.read()))
else:
image = sess.run(tf.image.decode_jpeg(img.read()))
height = image.shape[0]
width = image.shape[1]
tf.logging.info('Image size: %dx%d' % (width, height))
with tf.Graph().as_default():
with tf.Session().as_default() as sess:
# Read image data.
image_preprocessing_fn, _ = preprocessing_factory.get_preprocessing(
FLAGS.loss_model, is_training=False)
image = reader.get_image(FLAGS.image_file, height, width,
image_preprocessing_fn)
# Add batch dimension
image = tf.expand_dims(image, 0)
generated = model.net(image, training=False)
generated = tf.cast(generated, tf.uint8)
# Remove batch dimension
generated = tf.squeeze(generated, [0])
# Restore model variables.
saver = tf.train.Saver(tf.global_variables(),
write_version=tf.train.SaverDef.V1)
sess.run([
tf.global_variables_initializer(),
tf.local_variables_initializer()
])
# Use absolute path
FLAGS.model_file = os.path.abspath(FLAGS.model_file)
saver.restore(sess, FLAGS.model_file)
# Make sure 'generated' directory exists.
# generated_file = 'generated/res.jpg'
generated_file = FLAGS.image_file.replace("content", "res")
if os.path.exists('generated') is False:
os.makedirs('generated')
# Generate and write image data to file.
with open(generated_file, 'wb') as img:
start_time = time.time()
img.write(sess.run(tf.image.encode_jpeg(generated)))
end_time = time.time()
tf.logging.info('Elapsed time: %fs' % (end_time - start_time))
tf.logging.info('Done. Please check %s.' % generated_file)
#tf.train.export_meta_graph(filename='model.meta',collection_list=['input_tensor'],as_text=True)
tf.train.export_meta_graph(filename='model.meta', as_text=True)
def gen(imagename, png = False):
# Get image's height and width.
height = 0
width = 0
with open(imagename, 'rb') as img:
with tf.Session().as_default() as sess:
if png:
image = sess.run(tf.image.decode_png(img.read()))
else:
image = sess.run(tf.image.decode_jpeg(img.read()))
height = image.shape[0]
width = image.shape[1]
tf.logging.info('Image size: %dx%d' % (width, height))
with tf.Graph().as_default():
with tf.Session().as_default() as sess:
# Read image data.
image_preprocessing_fn, _ = preprocessing_factory.get_preprocessing(
FLAGS.loss_model,
is_training=False)
image = reader.get_image(imagename, height, width, image_preprocessing_fn)
# Add batch dimension
image = tf.expand_dims(image, 0)
generated = model.net(image, training=False)
generated = tf.cast(generated, tf.uint8)
# Remove batch dimension
generated = tf.squeeze(generated, [0])
# Restore model variables.
saver = tf.train.Saver(tf.global_variables(), write_version=tf.train.SaverDef.V1)
sess.run([tf.global_variables_initializer(), tf.local_variables_initializer()])
# Use absolute path
FLAGS.model_file = os.path.abspath(FLAGS.model_file)
saver.restore(sess, FLAGS.model_file)
# Make sure 'generated' directory exists.
generated_file = 'generated/' + imagename.split("/")[-2]+'/'+imagename[imagename.rfind("/")+1:]
print(imagename)
print(generated_file)
if os.path.exists('generated') is False:
os.makedirs('generated')
if os.path.exists( 'generated/' + imagename.split("/")[-2]) is False:
os.makedirs('generated/' + imagename.split("/")[-2])
# Generate and write image data to file.
with open(generated_file, 'wb') as img:
start_time = time.time()
img.write(sess.run(tf.image.encode_png(generated)))
end_time = time.time()
tf.logging.info('Elapsed time: %fs' % (end_time - start_time))
tf.logging.info('Done. Please check %s.' % generated_file)
def generation(image_file, model_file, generated_file):
result = dict()
# Get image's height and width.
height = 0
width = 0
with open(image_file, 'rb') as img:
with tf.Session().as_default() as sess:
if image_file.lower().endswith('png'):
image = sess.run(tf.image.decode_png(img.read()))
else:
image = sess.run(tf.image.decode_jpeg(img.read()))
height = image.shape[0]
width = image.shape[1]
print('Image size: %dx%d' % (width, height))
with tf.Graph().as_default():
with tf.Session().as_default() as sess:
# Read image data.
image_preprocessing_fn, _ = preprocessing_factory.get_preprocessing(
loss_model, is_training=False)
image = reader.get_image(image_file, height, width,
image_preprocessing_fn)
# Add batch dimension
image = tf.expand_dims(image, 0)
generated = model.net(image, training=False)
generated = tf.cast(generated, tf.uint8)
# Remove batch dimension
generated = tf.squeeze(generated, [0])
# Restore model variables.
saver = tf.train.Saver(tf.global_variables(),
write_version=tf.train.SaverDef.V1)
sess.run([
tf.global_variables_initializer(),
tf.local_variables_initializer()
])
# Use absolute path
model_file = os.path.abspath(model_file)
saver.restore(sess, model_file)
# Generate and write image data to file.
with open(generated_file, 'wb') as img:
start_time = time.time()
img.write(sess.run(tf.image.encode_jpeg(generated)))
end_time = time.time()
result['time'] = (end_time - start_time)
result['path'] = generated_file
return result
def main(models, img_c):
#tf.app.flags.DEFINE_string("image_file",img_c, "")
#FLAGS = tf.app.flags.FLAGS
FLAGS = arguement('vgg_16', models, img_c)
# Get image's height and width.
height = 0
width = 0
with open(FLAGS.image_file, 'rb') as img:
with tf.Session().as_default() as sess:
if FLAGS.image_file.lower().endswith('png'):
image = sess.run(tf.image.decode_png(img.read()))
else:
image = sess.run(tf.image.decode_jpeg(img.read()))
height = image.shape[0]
width = image.shape[1]
tf.logging.info('Image size: %dx%d' % (width, height))
with tf.Graph().as_default():
with tf.Session().as_default() as sess:
# Read image data.
image_preprocessing_fn, _ = preprocessing_factory.get_preprocessing(
FLAGS.loss_model, is_training=False)
image = reader.get_image(FLAGS.image_file, height, width,
image_preprocessing_fn)
# Add batch dimension
image = tf.expand_dims(image, 0)
generated = model.net(image, training=False)
generated = tf.cast(generated, tf.uint8)
# Remove batch dimension
generated = tf.squeeze(generated, [0])
# Restore model variables.
saver = tf.train.Saver(tf.global_variables(),
write_version=tf.train.SaverDef.V1)
sess.run([
tf.global_variables_initializer(),
tf.local_variables_initializer()
])
# Use absolute path
FLAGS.model_file = os.path.abspath(FLAGS.model_file)
saver.restore(sess, FLAGS.model_file)
# Make sure 'generated' directory exists.
generated_file = 'generated/res.jpg'
if os.path.exists('generated') is False:
os.makedirs('generated')
# Generate and write image data to file.
with open(generated_file, 'wb') as img:
#start_time = time.time()
img.write(sess.run(tf.image.encode_jpeg(generated)))
def main(_):
#获取图片的长宽
height = 0
width = 0
with open(FLAGS.image_file, 'rb') as img:
with tf.Session().as_default() as sess:
if FLAGS.image_file.lower().endswith('png'):
image = sess.run(tf.image.decode_png(img.read()))
else:
image = sess.run(tf.image.decode_jpeg(img.read()))
height = image.shape[0]
width = image.shape[1]
tf.logging.info('图片尺寸为: %dx%d' % (width, height))
with tf.Graph().as_default():
with tf.Session().as_default() as sess:
#读取图片信息
image_preprocessing_fn, _ = preprocessing_factory.get_preprocessing(
FLAGS.loss_model, is_training=False)
image = reader.get_image(FLAGS.image_file, height, width,
image_preprocessing_fn)
#增加维度
image = tf.expand_dims(image, 0)
generated = model.net(image, training=False)
generated = tf.cast(generated, tf.uint8)
#从tensor中删除所有大小是1的维度
generated = tf.squeeze(generated, [0])
#保存模型
saver = tf.train.Saver(tf.global_variables(),
write_version=tf.train.SaverDef.V1)
sess.run([
tf.global_variables_initializer(),
tf.local_variables_initializer()
])
"""获取已训练好的model"""
FLAGS.model_file = os.path.abspath(FLAGS.model_file)
saver.restore(sess, FLAGS.model_file)
"""生成转换style后的image"""
# 确认'generated'文件夹存在.
generated_file = 'generated/res.jpg'
if os.path.exists('generated') is False:
os.makedirs('generated')
# 将转换后的图片信息写入文件
with open(generated_file, 'wb') as img:
start_time = time.time()
img.write(sess.run(tf.image.encode_jpeg(generated)))
end_time = time.time()
tf.logging.info('Elapsed time: %fs' % (end_time - start_time))
tf.logging.info('Done. Please check %s.' % generated_file)
def main(_):
# Get image's height and width.
height = 0
width = 0
with open(FLAGS.image_file, 'rb') as img:
with tf.Session().as_default() as sess:
if FLAGS.image_file.lower().endswith('png'):
image = sess.run(tf.image.decode_png(img.read()))
else:
image = sess.run(tf.image.decode_jpeg(img.read()))
height = image.shape[0]
width = image.shape[1]
tf.logging.info('Image size %d X %d' % (width, height))
with tf.Graph().as_default():
with tf.Session().as_default() as sess:
#Read image data
image_preprocessing_fn, _ = preprocessing_factory.get_preprocessing(
FLAGS.loss_model, is_training=False)
image = reader.get_image(FLAGS.image_file, height, width,
image_preprocessing_fn)
# Add batch dimension
image = tf.expand_dims(image, 0)
generated = model.net(image, training=False)
generated = tf.cast(generated, tf.uint8)
# Remove batch dimension
generated = tf.squeeze(generated, [0])
# Restore model variables.
saver = tf.train.Saver(tf.global_variables(),
write_version=tf.train.SaverDef.V1)
sess.run([
tf.global_variables_initializer(),
tf.local_variables_initializer()
])
print('load model parameters...')
FLAGS.model_file = os.path.abspath(FLAGS.model_file)
saver.restore(sess, FLAGS.model_file)
# Make sure 'generated' directory exists.
generated_file = 'generated/res.jpg'
if os.path.exists('generated') is False:
os.mkdir('generated')
# Generate and write image data to file.
with open(generated_file, 'wb') as img:
start_time = time.time()
img.write(sess.run(tf.image.encode_jpeg(generated)))
end_time = time.time()
tf.logging.info('Elapsed time %fs' % (end_time - start_time))
tf.logging.info('Done. please check %s.' % generated_file)
def get_content_features(content_path, content_layers):
with tf.Graph().as_default() as g:
image = tf.expand_dims(reader.get_image(content_path, FLAGS.IMAGE_SIZE), 0)
net, _ = vgg.net(FLAGS.VGG_PATH, image)
layers = []
for layer in content_layers:
layers.append(net[layer])
with tf.Session() as sess:
return sess.run(layers + [image])
def get_style_features(style_paths, style_layers):
with tf.Graph().as_default() as g:
images = tf.stack([reader.get_image(style_paths,256)])
net, _ = vgg.net(images - reader.mean_pixel)
features = []
for layer in style_layers:
features.append(gram(net[layer]))
with tf.Session() as sess:
return sess.run(features)
def get_content_features(content_path, content_layers):
with tf.Graph().as_default() as g:
image = tf.expand_dims(reader.get_image(content_path, FLAGS.IMAGE_SIZE), 0)
net, _ = vgg.net(FLAGS.VGG_PATH, image)
layers = []
for layer in content_layers:
layers.append(net[layer])
with tf.Session() as sess:
return sess.run(layers + [image])
def gen_single():
""" Transfer an image. """
content_images = reader.get_image(FLAGS.content_image, FLAGS.image_size)
images = tf.stack([content_images])
generated_images = model.net(images / 255., if_train=False)
output_format = tf.saturate_cast(generated_images + reader.mean_pixel,
tf.uint8)
# Output path
model_path = os.path.join('models',
FLAGS.model_name + utils.get_model_suffix())
### model_p = model_p if not model_p.endswith("/") else model_p[:-1]
### model_p = os.path.split(model_p)
output_path = os.path.join("output",
FLAGS.model_name + utils.get_model_suffix())
if not os.path.exists(output_path):
os.makedirs(output_path)
with tf.Session() as sess:
file_ = tf.train.latest_checkpoint(model_path)
if not file_:
print('Could not find trained model in {}'.format(model_path))
return
print('Using model from {}'.format(file_))
# Get trained step
index = file_.rfind("-")
trained_step = file_[index:]
saver = tf.train.Saver()
saver.restore(sess, file_)
print("Style image:", FLAGS.content_image)
start_time = time.time()
# Run inference
images_t = sess.run(output_format)
elapsed = time.time() - start_time
print('Time: {}'.format(elapsed))
out_path = os.path.join(
output_path,
FLAGS.output + trained_step + '-' + str(int(time.time())) + '.jpg')
print("Save result in: ", out_path)
misc.imsave(out_path, images_t[0])
print('------------------------------------')
print('Finished!')
return
def get_style_features(style_paths, style_layers):
with tf.Graph().as_default() as g:
size = int(round(FLAGS.IMAGE_SIZE * FLAGS.STYLE_SCALE))
images = tf.pack([reader.get_image(path, size) for path in style_paths])
net, _ = vgg.net(FLAGS.VGG_PATH, images)
features = []
for layer in style_layers:
features.append(gram(net[layer]))
with tf.Session() as sess:
return sess.run(features)
def get_style_features(style_paths, style_layers):
with tf.Graph().as_default() as g:
size = int(round(FLAGS.IMAGE_SIZE * FLAGS.STYLE_SCALE))
images = tf.pack([reader.get_image(path, size) for path in style_paths])
net, _ = vgg.net(FLAGS.VGG_PATH, images)
features = []
for layer in style_layers:
features.append(gram(net[layer]))
with tf.Session() as sess:
return sess.run(features)
def get_style_features(style_paths, style_layers, net_type):
with tf.Graph().as_default() as g:
size = int(round(FLAGS.image_size * FLAGS.style_scale))
images = tf.stack(
[reader.get_image(path, size) for path in style_paths])
net, _ = vgg.net(FLAGS.vgg_path, images, net_type)
features = []
for layer in style_layers:
features.append(model.gram(net[layer], FLAGS.batch_size))
with tf.Session() as sess:
return sess.run(features)
def main(_):
# Get image's height and width.
height = 0
width = 0
with open(FLAGS.image_file, 'rb') as img:
with tf.Session().as_default() as sess:
if FLAGS.image_file.lower().endswith('png'):
image = sess.run(tf.image.decode_png(img.read()))
else:
image = sess.run(tf.image.decode_jpeg(img.read()))
height = image.shape[0]
width = image.shape[1]
tf.logging.info('Image size: %dx%d' % (width, height))
with tf.Graph().as_default():
with tf.Session().as_default() as sess:
# Read image data.
image_preprocessing_fn, _ = preprocessing_factory.get_preprocessing(
FLAGS.loss_model,
is_training=False)
image = reader.get_image(FLAGS.image_file, height, width, image_preprocessing_fn)
# Add batch dimension
image = tf.expand_dims(image, 0)
generated = model.net(image, training=False)
generated = tf.cast(generated, tf.uint8)
# Remove batch dimension
generated = tf.squeeze(generated, [0])
# Restore model variables.
saver = tf.train.Saver(tf.global_variables(), write_version=tf.train.SaverDef.V1)
sess.run([tf.global_variables_initializer(), tf.local_variables_initializer()])
# Use absolute path
FLAGS.model_file = os.path.abspath(FLAGS.model_file)
saver.restore(sess, FLAGS.model_file)
# Make sure 'generated' directory exists.
generated_file = 'generated/res.jpg'
if os.path.exists('generated') is False:
os.makedirs('generated')
# Generate and write image data to file.
with open(generated_file, 'wb') as img:
start_time = time.time()
img.write(sess.run(tf.image.encode_jpeg(generated)))
end_time = time.time()
tf.logging.info('Elapsed time: %fs' % (end_time - start_time))
tf.logging.info('Done. Please check %s.' % generated_file)
def get_style_features(style_paths, style_layers, image_size, style_scale,
vgg_path):
with tf.Graph().as_default(), tf.Session() as sess:
size = int(round(image_size * style_scale))
images = tf.stack(
[reader.get_image(path, size) for path in style_paths])
net, _ = vgg.net(vgg_path, images - vgg.MEAN_PIXEL)
features = []
for layer in style_layers:
features.append(gram(net[layer]))
return sess.run(features)
def freeze_model():
""" freeze graph. """
input_node_names = "input_node"
output_node_names = "output_node"
content_images = reader.get_image(FLAGS.content_image, FLAGS.image_size)
images = tf.pack([content_images])
input_images = tf.placeholder(dtype=tf.float32, name=input_node_names)
generated_images = model.net(input_images / 255., if_train=False)
output_format = tf.saturate_cast(generated_images + reader.mean_pixel,
tf.uint8,
name=output_node_names)
with tf.Session() as sess:
checkpoint_file = tf.train.latest_checkpoint(FLAGS.model)
if not checkpoint_file:
print('Could not find trained model in {}'.format(FLAGS.model))
return
print('Using model from {}'.format(checkpoint_file))
saver = tf.train.Saver()
saver.restore(sess, checkpoint_file)
in_images = sess.run(images)
images_t = sess.run(output_format, feed_dict={input_images: in_images})
# Save graph
tf.train.write_graph(sess.graph.as_graph_def(), FLAGS.model,
FLAGS.in_graph_name)
checkpoint_prefix = os.path.join(FLAGS.model, "saved_checkpoint")
checkpoint_state_name = "checkpoint_state"
# We save out the graph to disk, and then call the const conversion
# routine.
input_graph_path = os.path.join(FLAGS.model, FLAGS.in_graph_name)
input_saver_def_path = ""
input_binary = False
restore_op_name = "save/restore_all"
filename_tensor_name = "save/Const:0"
output_graph_path = os.path.join(FLAGS.model, FLAGS.out_graph_name)
clear_devices = False
freeze_graph.freeze_graph(input_graph_path, input_saver_def_path,
input_binary, checkpoint_file, output_node_names,
restore_op_name, filename_tensor_name,
output_graph_path, clear_devices, "")
print('------------------------------------')
print('Finished!')
def style_transform(style, model_file, img_file, img_filename, result_file):
height = 0
width = 0
with open(img_file, 'rb') as img:
img_data = img
with tf.Session().as_default() as sess:
if img_file.lower().endswith('png'):
image = sess.run(tf.image.decode_png(img.read()))
else:
image = sess.run(tf.image.decode_jpeg(img.read()))
height = image.shape[0]
width = image.shape[1]
print('Image size: %dx%d' % (width, height))
with tf.Graph().as_default():
with tf.Session().as_default() as sess:
image_preprocessing_fn, _ = preprocessing_factory.get_preprocessing(
FLAGS.loss_model, is_training=False)
image = reader.get_image(img_file, height, width,
image_preprocessing_fn)
image = tf.expand_dims(image, 0)
generated = model.net(image, training=False)
generated = tf.cast(generated, tf.uint8)
generated = tf.squeeze(generated, [0])
saver = tf.train.Saver(tf.global_variables())
sess.run([
tf.global_variables_initializer(),
tf.local_variables_initializer()
])
FLAGS.model_file = os.path.abspath(model_file)
saver.restore(sess, FLAGS.model_file)
generated_file = 'static/img/generated/' + result_file
if os.path.exists('static/img/generated') is False:
os.makedirs('static/img/generated')
with open(generated_file, 'wb') as img:
result_data = img
start_time = time.time()
img.write(sess.run(tf.image.encode_jpeg(generated)))
end_time = time.time()
now = datetime.datetime.now()
print('Elapsed time: %fs' % (end_time - start_time))
print('Done. Please check %s.' % generated_file)
cluster = Cluster(contact_points=['127.0.0.1'], port=9042)
session = cluster.connect()
log.info("setting keyspace...")
session.set_keyspace(KEYSPACE)
session.execute(
"""
INSERT INTO imagerecord (time,style,input_name,output_name)
VALUES (%s,%s,%s,%s)
""", (now, style, img_filename, result_file))
def get_content_features(content_layers):
with tf.Graph().as_default() as g:
#tf.expand_dims 在位置0扩展了一个维度[2,3]=>[1,2,3]变成图片
image = tf.expand_dims(reader.get_image(content_path, IMAGE_SIZE), 0)
net, _ = vgg.net(image)
layers = []
for layer in content_layers:
# net字典形式，net{"conv1_1":"conv1_1的结果“}
layers.append(net[layer])
with tf.Session() as sess:
#返回features + 图片
return sess.run(layers + [image])
def get_style_features(style_paths, style_layers):
with tf.Graph().as_default() as g:
size = int(round(FLAGS.IMAGE_SIZE * FLAGS.STYLE_SCALE))
images = tf.pack(
[reader.get_image(path, size) for path in style_paths])
net, _ = vgg.net(FLAGS.VGG_PATH, images)
features = []
for layer in style_layers:
features.append(gram(net[layer]))
config = tf.ConfigProto()
config.gpu_options.per_process_gpu_memory_fraction = 1.0
with tf.Session(config=config) as sess:
return sess.run(features)
def get_style_features(style_paths, style_layers):
with tf.Graph().as_default() as g:
size = int(round(IMAGE_SIZE))
images = tf.stack(
[reader.get_image(path, size) for path in style_paths])
net, _ = vgg.net(images - reader.mean_pixel)
features = []
for layer in style_layers:
features.append(gram(net[layer]))
with tf.Session() as sess:
return sess.run(features)
def neualstyle(model_file, image_file):
# 初始化参数
loss_model = 'vgg_16'
height = 0
width = 0
# 读取图片，并将图片转化为tensor张量
with open(image_file, 'rb') as img:
with tf.Session().as_default() as sess:
if image_file.lower().endswith('png'):
# 将图片转化为张量，tensor，[height,width,channels]
image = sess.run(tf.image.decode_png(img.read()))
else:
image = sess.run(tf.image.decode_jpeg(img.read()))
height = image.shape[0]
width = image.shape[1]
with tf.Graph().as_default():
with tf.Session().as_default() as sess:
# 获取处理图片的方法
image_preprocessing_fn, _ = preprocessing_factory.get_preprocessing(
loss_model, is_training=False)
# 获取预处理后的tensor
image = reader.get_image(image_file, height, width,
image_preprocessing_fn)
# 添加一个维度，oldshape=[height,width,3] ——> newshape=[1,height,width,3]
image = tf.expand_dims(image, 0)
# 获取神经网络
generated = model.net(image, training=False)
generated = tf.cast(generated, tf.uint8)
# 去掉增加的维
generated = tf.squeeze(generated, [0])
# 保存模型的变量
saver = tf.train.Saver(tf.global_variables(),
write_version=tf.train.SaverDef.V1)
sess.run([
tf.global_variables_initializer(),
tf.local_variables_initializer()
])
saver.restore(sess, model_file)
# 保存图片
generated_name = image_file.split('/')[-1]
generated_file = os.getcwd(
) + '/wx_robot/handler/neuralstyle/neuralpic/' + generated_name
with open(generated_file, 'wb') as img:
img.write(sess.run(tf.image.encode_jpeg(generated)))
def main(_): #main(_)表示什么意思？？？
#get image width and height
width = 0
height = 0
with open(FLAGS.image_file,'rb') as img:
with tf.Session().as_default as sess:
if FLAGS.image_file.lower().endswith('png'):
image = sess.run(tf.image.decode_png(img.read()))
else:
image = sess.run(tf.image.decode_jpge(img.read()))
height = image.shape[0]
width = image.shape[1]
tf.logging.info('image size is : %dx%d' % (height,width))
with tf.Graph().as_default():
with tf.Session.as_default() as sess:
#Read image data
image_preprocessing_fn,_ = preprocessing_factory.get_preprocessing(FLAGS.loss_model,is_training = False) #FLAGS.loss_model为用于评估的model，该function返回的是，在loss_model下，对image进行预处理的function
image = reader.get_image(FLAGS.image_file,height,width,image_preprocessing_fn) #利用image_preprocessing_fn函数，对image进行预处理，返回预处理后的image
#Add batch dimension
image = np.expand_dims(image,0)
generated = model.net(image,training = False) #形成图像风格迁移：返回的是一个需要sess.run()的model
generated = tf.cast(generated,tf.uint8) #将generated的数据格式转为uint8
#remove batch dimension
generated = np.squeeze(generated,[0])
#restore model variables??? 预训练模型，为什么还要在刷新其参数？？？
saver = tf.train.saver(tf.global_variables(),write_version = tf.train.SaverDef.V1) #创建类实例
#是否可去？？？： sess.run([tf.global_variables_initializer(),tf.local_variables_initializer()]) #这里运行的是model.net中的图像生成网络？input image在最开始的时候，已经输入图像生成网络，这里是进行训练，但是，在model.net(train=False)，所以不需要进行训练，那么这里为什么还有在对参数进行初始化呢？不合逻辑，正常理解，应该是，重载模型参数，saver.restore()，然后，在用sess.run(generated)???
#use absolute path
FLAGS.model_file = os.path.abspath(FLAGS.model_file)
saver.restore(sess,FLAGS.model_file) #重载模型参数，为什么要重载模型参数，不是已经风格迁移了吗？？？
#上边这几句是不是写的反了？？？
#make sure generated directory exists
generated_file = 'generated/res.jpg'
if not os.path.exists('generated'):
os.makedirs('generated')
#generate and write image data to file
with open(generated_file,'wb') as img:
start_time = time.time()
img.write(sess.run(tf.encode_jpeg(generated)))
end_time = time.time()
tf.logging.info('Elapsed time: %fs' % (end_time - start_time))
tf.logging.info('Done. Please check %s.' % generated_file)
def generate():
height, width = 0, 0
with open((FLAGS.image_path + FLAGS.input_image), 'rb') as img:
with tf.Session().as_default() as sess:
if FLAGS.input_image.lower().endswith('png'):
image = sess.run(tf.image.decode_png(img.read()))
else:
image = sess.run(tf.image.decode_jpeg(img.read()))
height = image.shape[0]
width = image.shape[1]
tf.logging.info('Image size: %dx%d' % (width, height))
with tf.Graph().as_default():
with tf.Session().as_default() as sess:
img_prep_fn, _ = preprocessing_factory.get_preprocessing(
FLAGS.loss_model, is_training=False)
img = reader.get_image((FLAGS.image_path + FLAGS.input_image),
height, width, img_prep_fn)
# Add batch dimension
img = tf.expand_dims(img, 0)
gen_img = tf.cast(model.img_trans_net(img, training=False),
tf.uint8)
# Remove batch dimension
gen_img = tf.squeeze(gen_img, [0])
# Restore model variables.
saver = tf.train.Saver(tf.global_variables(),
write_version=tf.train.SaverDef.V2)
sess.run([
tf.global_variables_initializer(),
tf.local_variables_initializer()
])
# Use absolute path
FLAGS.model_file = os.path.abspath(FLAGS.model_path +
FLAGS.model_file)
saver.restore(sess, FLAGS.model_file)
# Make sure FLAGS.image_path directory exists.
gen_file = FLAGS.image_path + FLAGS.output_image
if os.path.exists(FLAGS.image_path) is False:
os.makedirs(FLAGS.image_path)
# generate image
with open(gen_file, 'wb') as img:
start = time.time()
img.write(sess.run(tf.image.encode_jpeg(gen_img)))
elapsed = time.time() - start
print('Elapsed time: {}s'.format(elapsed))
tf.logging.info('Elapsed time: {}s'.format(elapsed))
tf.logging.info('Done. Please check {}.'.format(gen_file))
def infer(self,model_file,image_file):
image_file=decode_img(image_file)
image1 = cv2.imread(image_file)
height = image1.shape[0]
width = image1.shape[1]
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
print('Image size: %dx%d' % (width, height))
with tf.Graph().as_default():
with tf.Session(config=config).as_default() as sess:
# Read image data.
image_preprocessing_fn, _ = preprocessing_factory.get_preprocessing(
self.loss_model,
is_training=False)
image = reader.get_image(image_file, height, width, image_preprocessing_fn)
# Add batch dimension
image = tf.expand_dims(image, 0)
generated = model.net(image, training=False)
generated = tf.cast(generated, tf.uint8)
# Remove batch dimension
generated = tf.squeeze(generated, [0])
# Restore model variables.
saver = tf.train.Saver(tf.global_variables(), write_version=tf.train.SaverDef.V1)
sess.run([tf.global_variables_initializer(), tf.local_variables_initializer()])
# Use absolute path
model_file = os.path.abspath(model_file)
saver.restore(sess, model_file)
# Make sure 'generated' directory exists.
generated_file ='./tmp/'+time.strftime('%Y_%m_%d_%H_%M_%S',time.localtime(time.time()))+'.jpg'
if os.path.exists('tmp') is False:
os.makedirs('tmp')
# Generate and write image data to file.
with open(generated_file, 'wb') as img:
start_time = time.time()
img.write(sess.run(tf.image.encode_jpeg(generated)))
end_time = time.time()
print('Elapsed time: %fs' % (end_time - start_time))
os.remove(image_file)
with open(generated_file,'rb') as f:
style_base64=base64.b64encode(f.read())
s = style_base64.decode()
#str(style_base64, encoding = "utf8")
return s
def main(argv=None):
network_fn = nets_factory.get_network_fn('vgg_16',
num_classes=1,
is_training=False)
image_preprocessing_fn, image_unprocessing_fn = preprocessing_factory.get_preprocessing(
'vgg_16', is_training=False)
preprocess_content_image = reader.get_image(FLAGS.CONTENT_IMAGE,
FLAGS.IMAGE_SIZE)
# add bath for vgg net training
preprocess_content_image = tf.expand_dims(preprocess_content_image, 0)
_, endpoints_dict = network_fn(preprocess_content_image,
spatial_squeeze=False)
# Log the structure of loss network
tf.logging.info(
'Loss network layers(You can define them in "content_layers" and "style_layers"):'
)
for key in endpoints_dict:
tf.logging.info(key)
"""Build Losses"""
# style_features_t = losses.get_style_features(endpoints_dict, FLAGS.STYLE_LAYERS)
content_loss, generaged_image = losses.content_loss(
endpoints_dict, FLAGS.CONTENT_LAYERS, FLAGS.CONTENT_IMAGE)
style_loss, style_loss_summary = losses.style_loss(endpoints_dict,
FLAGS.style_layers,
FLAGS.STYLE_IMAGE)
tv_loss = losses.total_variation_loss(
generaged_image) # use the unprocessed image
loss = FLAGS.STYLE_WEIGHT * style_loss + FLAGS.CONTENT_WEIGHT * content_loss + FLAGS.TV_WEIGHT * tv_loss
train_op = tf.train.AdamOptimizer(FLAGS.LEARNING_RATE).minimize(loss)
output_image = tf.image.encode_png(
tf.saturate_cast(
tf.squeeze(generaged_image) + reader.mean_pixel, tf.uint8))
with tf.Session() as sess:
sess.run(tf.initialize_all_variables())
start_time = time.time()
for step in range(FLAGS.NUM_ITERATIONS):
_, loss_t, cl, sl = sess.run(
[train_op, loss, content_loss, style_loss])
elapsed = time.time() - start_time
start_time = time.time()
print(step, elapsed, loss_t, cl, sl)
image_t = sess.run(output_image)
with open('out.png', 'wb') as f:
f.write(image_t)
def get_style_features():
with tf.Graph().as_default() as g:
size = int(round(IMAGE_SIZE))
images = tf.stack([reader.get_image(style_path, size)])
net, _ = vgg.net(images)
features = []
for layer in STYLE_LAYERS:
# net字典形式，net{"conv1_1":"1_1的结果“}
# layer 依次 = "relu1_1,relu2_1,relu3_1,relu4_1,relu5_1"
#返回每一层的gram矩阵
features.append(gram(net[layer]))
with tf.Session() as sess:
return sess.run(features)
def main(_):
height = 0
width = 0
with open(FLAGS.image_file, 'rb') as img:
with tf.Session().as_default() as sess:
if FLAGS.image_file.lower().endswith('png'):
image = sess.run(tf.image.decode_png(img.read()))
else:
image = sess.run(tf.image.decode_jpeg(img.read()))
height = image.shape[0]
width = image.shape[1]
tf.logging.info('Image size: %dx%d' % (width, height))
with tf.Graph().as_default():
with tf.Session().as_default() as sess:
image_preprocessing_fn, _ = preprocessing_factory.get_preprocessing(
FLAGS.loss_model, is_training=False)
image = reader.get_image(FLAGS.image_file, height, width,
image_preprocessing_fn)
image = tf.expand_dims(image, 0)
generated = model.net(image, training=False)
generated = tf.squeeze(generated, [0])
#saver = tf.train.Saver(tf.all_variables())
#sess.run([tf.initialize_all_variables(), tf.initialize_local_variables()])
saver = tf.train.Saver(tf.global_variables())
sess.run([
tf.global_variables_initializer(),
tf.local_variables_initializer()
])
FLAGS.model_file = os.path.abspath(FLAGS.model_file)
saver.restore(sess, FLAGS.model_file)
start_time = time.time()
generated = sess.run(generated)
generated = tf.cast(generated, tf.uint8)
end_time = time.time()
tf.logging.info('Elapsed time: %fs' % (end_time - start_time))
filename_image, file_extension = os.path.splitext(
os.path.basename(FLAGS.image_file))
filename_model, tmp_extension = os.path.splitext(
os.path.basename(FLAGS.model_file))
generated_file = 'generated/' + filename_image + "-" + filename_model + file_extension
if os.path.exists('generated') is False:
os.makedirs('generated')
with open(generated_file, 'wb') as img:
img.write(sess.run(tf.image.encode_jpeg(generated)))
tf.logging.info('Done. Please check %s.' % generated_file)
def img(image_file='img/test.jpg',
loss_model='vgg_16',
image_size=256,
model_file='models/wave.ckpt-done'):
height = 0
width = 0
with open(image_file, 'rb') as img:
with tf.Session().as_default() as sess:
if image_file.lower().endswith('png'):
image = sess.run(tf.image.decode_png(img.read()))
else:
image = sess.run(tf.image.decode_jpeg(img.read()))
height = image.shape[0]
width = image.shape[1]
# tf.logging.info('Image size: %dx%d' % (width, height))
with tf.Graph().as_default():
with tf.Session().as_default() as sess:
# Read image data.
image_preprocessing_fn, _ = preprocessing_factory.get_preprocessing(
loss_model, is_training=False)
image = reader.get_image(image_file, height, width,
image_preprocessing_fn)
# Add batch dimension
image = tf.expand_dims(image, 0)
generated = model.net(image, training=False)
generated = tf.cast(generated, tf.uint8)
# Remove batch dimension
generated = tf.squeeze(generated, [0])
# Restore model variables.
saver = tf.train.Saver(tf.global_variables(),
write_version=tf.train.SaverDef.V1)
sess.run([
tf.global_variables_initializer(),
tf.local_variables_initializer()
])
# Use absolute path
model_file = os.path.abspath(model_file)
saver.restore(sess, model_file)
# Make sure 'generated' directory exists.
generated_file = 'generated/test3.jpg'
if os.path.exists('generated') is False:
os.makedirs('generated')
# Generate and write image data to file.
with open(generated_file, 'wb') as img:
# start_time = time.time()
img.write(sess.run(tf.image.encode_jpeg(generated)))
def get_style_feature():
with tf.Graph().as_default():
preprocess_func, unprocess_func = preprocessing.preprocessing_factory.get_preprocessing(
args.loss_model, is_training=False)
style_img = reader.get_image(args.style_dir, args.size, args.size,
preprocess_func)
style_img = tf.expand_dims(style_img, 0)
loss_model = nets.nets_factory.get_network_fn(args.loss_model,
1,
is_training=False)
_, end_dict = loss_model(style_img, spatial_squeeze=False)
init_loss_model = load_pretrained_weight(args.loss_model)
#init_loss_model2 = load_pretrained_weight(args.loss_model)
features = []
feature_grams = []
#sess.run([tf.global_variables_initializer() , tf.local_variables_initializer()])
#sess.run([init_loss_model])
for layer in loss_config.style_loss_dict[args.loss_model]:
#print('--layer--' + layer)
feature = end_dict[layer]
gram = losses.gram_matrix(feature)
feature_s = tf.squeeze(feature, [0])
gram_s = tf.squeeze(gram, [0])
#f , g = sess.run([ feature_s , gram_s ])
features.append(feature_s)
feature_grams.append(gram_s)
with tf.Session() as sess:
init_loss_model(sess)
ff, gg = sess.run([features, feature_grams])
#print('qwq')
return ff, gg
def main(_):
height = 0
width = 0
for head in demo_list:
model_file = 'models/%s/fast-style-model.ckpt-2000' % head
with open(FLAGS.image_file, 'rb') as img:
with tf.Session().as_default() as sess:
if FLAGS.image_file.lower().endswith('png'):
image = sess.run(tf.image.decode_png(img.read()))
else:
image = sess.run(tf.image.decode_jpeg(img.read()))
height = image.shape[0]
width = image.shape[1]
tf.logging.info('Image size: %dx%d' % (width, height))
with tf.Graph().as_default():
with tf.Session(config=tf.ConfigProto(
gpu_options=gpu_options)).as_default() as sess:
image_preprocessing_fn, _ = preprocessing_factory.get_preprocessing(
FLAGS.loss_model, is_training=False)
image = reader.get_image(FLAGS.image_file, height, width,
image_preprocessing_fn)
image = tf.expand_dims(image, 0)
generated = model.net(image, training=False)
generated = tf.squeeze(generated, [0])
saver = tf.train.Saver(tf.all_variables())
sess.run([
tf.initialize_all_variables(),
tf.initialize_local_variables()
])
#name = FLAGS.model_file.split('/')[1]
model_file = os.path.abspath(model_file)
saver.restore(sess, model_file)
start_time = time.time()
generated = sess.run(generated)
generated = tf.cast(generated, tf.uint8)
end_time = time.time()
tf.logging.info('Elapsed time: %fs' % (end_time - start_time))
generated_file = 'generated/result_0338/%s.jpg' % head
#generated_file = 'generated/%s.jpg'%head
if os.path.exists('generated') is False:
os.makedirs('generated')
with open(generated_file, 'wb') as img:
img.write(sess.run(tf.image.encode_jpeg(generated)))
tf.logging.info('Done. Please check %s.' % generated_file)
def get_content_features(content_layers):
with tf.Graph().as_default() as g:
#tf.expand_dims 在位置0扩展了一个维度[2,3]=>[1,2,3]变成图片
fea_content = []
for file in range(1000):
file_path = "GrayPics/"
file_name = str(file) + ".jpg"
img_path = path.join(file_path, file_name)
image = tf.stack([reader.get_image(img_path,256)])
net, _ = vgg.net(image)
for layer in content_layers:
# net字典形式，net{"conv1_1":"conv1_1的结果“}
fea_content.append(gram(net[layer]))
with tf.Session() as sess:
#返回features + 图片
return sess.run(fea_content)
def main(_):
height = 0
width = 0
with open(FLAGS.image_file, 'rb') as img:
with tf.Session().as_default() as sess:
if FLAGS.image_file.lower().endswith('png'):
image = sess.run(tf.image.decode_png(img.read()))
else:
image = sess.run(tf.image.decode_jpeg(img.read()))
height = image.shape[0]
width = image.shape[1]
tf.logging.info('Image size: %dx%d' % (width, height))
with tf.Graph().as_default():
with tf.Session().as_default() as sess:
image_preprocessing_fn, _ = preprocessing_factory.get_preprocessing(
FLAGS.loss_model,
is_training=False)
"""获取经过预处理的输入图片，用于后面获取图片的content"""
image = reader.get_image(FLAGS.image_file, height, width, image_preprocessing_fn)
image = tf.expand_dims(image, 0)
generated = model.transform_network(image, training=False)
generated = tf.squeeze(generated, [0])
saver = tf.train.Saver(tf.global_variables(), write_version=tf.train.SaverDef.V1)
sess.run([tf.global_variables_initializer(), tf.local_variables_initializer()])
"""获取已训练好的model"""
FLAGS.model_file = os.path.abspath(FLAGS.model_file)
saver.restore(sess, FLAGS.model_file)
"""生成转换style后的image"""
start_time = time.time()
generated = sess.run(generated)
generated = tf.cast(generated, tf.uint8)
end_time = time.time()
tf.logging.info('Elapsed time: %fs' % (end_time - start_time))
generated_file = FLAGS.target_file
if os.path.exists('static/img/generated') is False:
os.makedirs('static/img/generated')
with open(generated_file, 'wb') as img:
img.write(sess.run(tf.image.encode_jpeg(generated)))
tf.logging.info('Done. Please check %s.' % generated_file)
def style_transform(style, model_file, img_file, result_file):
height = 0
width = 0
with open(img_file, 'rb') as img:
with tf.Session().as_default() as sess:
if img_file.lower().endswith('png'):
image = sess.run(tf.image.decode_png(img.read()))
else:
image = sess.run(tf.image.decode_jpeg(img.read()))
height = image.shape[0]
width = image.shape[1]
print('Image size: %dx%d' % (width, height))
with tf.Graph().as_default():
with tf.Session().as_default() as sess:
image_preprocessing_fn, _ = preprocessing_factory.get_preprocessing(
FLAGS.loss_model,
is_training=False)
image = reader.get_image(img_file, height, width, image_preprocessing_fn)
image = tf.expand_dims(image, 0)
generated = model.transform_network(image, training=False)
generated = tf.squeeze(generated, [0])
saver = tf.train.Saver(tf.global_variables())
sess.run([tf.global_variables_initializer(), tf.local_variables_initializer()])
FLAGS.model_file = os.path.abspath(model_file)
saver.restore(sess, FLAGS.model_file)
start_time = time.time()
generated = sess.run(generated)
generated = tf.cast(generated, tf.uint8)
end_time = time.time()
print('Elapsed time: %fs' % (end_time - start_time))
generated_file = 'static/img/generated/' + result_file
if os.path.exists('static/img/generated') is False:
os.makedirs('static/img/generated')
with open(generated_file, 'wb') as img:
img.write(sess.run(tf.image.encode_jpeg(generated)))
print('Done. Please check %s.' % generated_file)
def main(_):
height = 0
width = 0
with open(FLAGS.image_file, 'rb') as img:
with tf.Session().as_default() as sess:
if FLAGS.image_file.lower().endswith('png'):
image = sess.run(tf.image.decode_png(img.read()))
else:
image = sess.run(tf.image.decode_jpeg(img.read()))
height = image.shape[0]
width = image.shape[1]
tf.logging.info('Image size: %dx%d' % (width, height))
with tf.Graph().as_default():
with tf.Session().as_default() as sess:
image_preprocessing_fn, _ = preprocessing_factory.get_preprocessing(
FLAGS.loss_model,
is_training=False)
image = reader.get_image(FLAGS.image_file, height, width, image_preprocessing_fn)
image = tf.expand_dims(image, 0)
generated = model.net(image, training=False)
generated = tf.squeeze(generated, [0])
saver = tf.train.Saver(tf.all_variables())
sess.run([tf.initialize_all_variables(), tf.initialize_local_variables()])
FLAGS.model_file = os.path.abspath(FLAGS.model_file)
saver.restore(sess, FLAGS.model_file)
start_time = time.time()
generated = sess.run(generated)
generated = tf.cast(generated, tf.uint8)
end_time = time.time()
tf.logging.info('Elapsed time: %fs' % (end_time - start_time))
generated_file = 'generated/res.jpg'
if os.path.exists('generated') is False:
os.makedirs('generated')
with open(generated_file, 'wb') as img:
img.write(sess.run(tf.image.encode_jpeg(generated)))
tf.logging.info('Done. Please check %s.' % generated_file)