Good job!
Do you plan open source training code?

这个产品瞬间就需要8GB的显示卡内存,需求真的是蛮高的.
segs,这个是什么文件？？

请问运行main.py的时候报这个错误是什么原因，求指教。

The paper mentions this portion here. I have 2 questions.

1. Which layer are you extracting the full face from the eyes, and hair to get the face like the results above without eyes and hair.
2. Which layer are you performing histogram matching from Y I^ref to I^src

Does anyone know where to find the face parsing code? I need this code for another dataset

When i use img_size = 512 i have error

ValueError: Cannot feed value of shape (1, 512, 512, 3) for Tensor 'X:0', which has shape '(?, 256, 256, 3)'
on the line:
Xs_ = sess.run(Xs, feed_dict={X: X_img, Y: Y_img})

may be need other model.meta?

can you give an example of the HM in makeup loss ?

as the title

Seen from your network graph, there is no conv bias before instance normalization?

Thank you for showing interesting papers and models.
By the way, I didn't see any code that used face_align in main.py, but I used that code and felt the model improved. Do you know it?

您好，能分享一下训练的代码吗？万分感谢

RT

我把img_size 改为512，程序运行报错，256时正确，请问如何解决？
def preprocess(img):
return (img / 511. - 0.5) * 2
batch_size = 1
img_size = 512
报错：
Traceback (most recent call last):
File "D:\steelsoft\makeup_artist\main.py", line 58, in
Xs_ = sess.run(Xs, feed_dict={X: X_img, Y: Y_img})
File "D:\steelsoft\makeup_artist\venv\lib\site-packages\tensorflow\python\client\session.py", line 968, in run
result = self._run(None, fetches, feed_dict, options_ptr,
File "D:\steelsoft\makeup_artist\venv\lib\site-packages\tensorflow\python\client\session.py", line 1165, in _run
raise ValueError(
ValueError: Cannot feed value of shape (1, 512, 512, 3) for Tensor X:0, which has shape (None, 256, 256, 3)

我的代码如下：
tf.compat.v1.disable_eager_execution()

parser = argparse.ArgumentParser()
parser.add_argument('--no_makeup', type=str, default=os.path.join('imgs', 'no_makeup', 'yuan_2.jpg'),
help='path to the no_makeup image')

args = parser.parse_args()

def preprocess(img):
return (img / 511. - 0.5) * 2

def deprocess(img):
return (img + 1) / 2

batch_size = 1
img_size = 512
no_makeup = cv2.resize(imread(args.no_makeup), (img_size, img_size))
print(args.no_makeup)
X_img = np.expand_dims(preprocess(no_makeup), 0)
makeups = glob.glob(os.path.join('imgs', 'makeup', '.'))

tf.reset_default_graph()

tf.compat.v1.reset_default_graph()

sess = tf.Session()

sess = tf.compat.v1.Session()

sess.run(tf.global_variables_initializer())

sess.run(tf.compat.v1.global_variables_initializer())

saver = tf.train.import_meta_graph(os.path.join('model', 'model.meta'))

saver = tf.compat.v1.train.import_meta_graph(os.path.join('model', 'model.meta'))
saver.restore(sess, tf.train.latest_checkpoint('model'))

graph = tf.get_default_graph()

graph = tf.compat.v1.get_default_graph()
X = graph.get_tensor_by_name('X:0')
Y = graph.get_tensor_by_name('Y:0')
Xs = graph.get_tensor_by_name('generator/xs:0')

for i in range(len(makeups)):
makeup = cv2.resize(imread(makeups[i]), (img_size, img_size))
Y_img = np.expand_dims(preprocess(makeup), 0)
Xs_ = sess.run(Xs, feed_dict={X: X_img, Y: Y_img})
Xs_ = deprocess(Xs_)
result = Xs_[0]

imsave('result.jpg', result)

This project seems to be very helpful to me. I have cloned the project but cannot download the model from baidu because it requires a Chinese account to do this task.
Can you please upload the model to Google Drive. I would very appreciate it!