About the Parsing net inference code

[kingsj0405]

Hello.
Thank you for your great work.
I'm currently try to use Face Parsing net of Lin et al. method to get face and hairear segmentation map like following image.

But the output of network is not desirable...
(+) I get (256, 256, 3) segmentation map although input image is (512, 512, 3).

My code for getting network is as following:

def get_pretrained_face_parser(sess=None):
    graph_input_img = tf.placeholder(dtype=tf.float32)
    graph_facial_landmark = tf.placeholder(dtype=tf.float32, shape=[1, 68, 2])
    graph_face_parsing = Parsing(graph_input_img, graph_facial_landmark)
    graph = tf.get_default_graph()
    if not sess:
        sess = tf.Session()
        sess.run(tf.global_variables_initializer())
    # get variable list
    var_list = tf.trainable_variables()
    g_list = tf.global_variables()
    bn_moving_vars = [g for g in g_list if 'moving_mean' in g.name]
    bn_moving_vars += [g for g in g_list if 'moving_variance' in g.name]
    var_list +=bn_moving_vars
    parser_vars = [v for v in var_list if 'FaceParser' in v.name]
    # restore weight provided by Lin et al.
    saver_parser = tf.train.Saver(var_list=parser_vars)
    saver_parser.restore(sess, os.path.join('./training/pretrained_weights/parsing_net','faceparser_public'))
    return sess, graph, graph_face_parsing, graph_input_img, graph_facial_landmark

Also I use Preprocess function from preprocess/preprocess_utils.py. Is there some suggestion to utilize your code?

I already check #19
Have a nice day.