test.py

import argparse

import cv2
import numpy as np
import tensorflow as tf
import neuralgym as ng

from inpaint_model import InpaintCAModel


parser = argparse.ArgumentParser()
parser.add_argument('--image', default='', type=str,
                    help='The filename of image to be completed.')
parser.add_argument('--mask', default='', type=str,
                    help='The filename of mask, value 255 indicates mask.')
parser.add_argument('--output', default='output.png', type=str,
                    help='Where to write output.')
parser.add_argument('--checkpoint_dir', default='', type=str,
                    help='The directory of tensorflow checkpoint.')


if __name__ == "__main__":
    FLAGS = ng.Config('inpaint.yml')
    args, unknown = parser.parse_known_args()

    model = InpaintCAModel()
    image = np.load(args.image)
    image *= 4096.

    mask = np.load(args.mask)
    mask *= 4096.

    assert image.shape == mask.shape

    image = np.stack([image,image,image],2)
    mask = np.stack([mask,mask,mask],2)

    h, w,_ = image.shape
    grid = 8
    image = image[:h//grid*grid, :w//grid*grid, :]
    mask = mask[:h//grid*grid, :w//grid*grid, :]

    image = np.expand_dims(image, 0)
    mask = np.expand_dims(mask, 0)
    input_image = np.concatenate([image, mask], axis=2)
    sess_config = tf.ConfigProto()
    sess_config.gpu_options.allow_growth = True
    with tf.Session(config=sess_config) as sess:
        input_image = tf.constant(input_image, dtype=tf.float32)
        output = model.build_server_graph(FLAGS, input_image)
        output = (output + 1.) * 2048
        output = tf.reverse(output, [-1])
        output = tf.saturate_cast(output, tf.float32)
        # load pretrained model
        vars_list = tf.get_collection(tf.GraphKeys.GLOBAL_VARIABLES)
        assign_ops = []
        for var in vars_list:
            vname = var.name
            from_name = vname
            var_value = tf.contrib.framework.load_variable(args.checkpoint_dir, from_name)
            assign_ops.append(tf.assign(var, var_value))
        sess.run(assign_ops)
        print('Model loaded.')
        result = sess.run(output)
        one_channel_img = result[0][:, :, ::-1][:,:,0]
        one_channel_img = np.true_divide(one_channel_img,4096)
        np.save(args.output.replace("png","npy"), one_channel_img)