model_filename = None
if len(sys.argv) >= 2:
start_iter = int(sys.argv[1])
model_filename = os.path.join(MODEL_DIR, 'model%d.ckpt' % start_iter)
# Generator code
K.set_learning_phase(1) # to accommodate Keras layers
G_Z = tf.placeholder(tf.float32,
shape=[None, IMAGE_SZ, IMAGE_SZ, 4],
name='G_Z')
DG_X = tf.placeholder(tf.float32,
shape=[None, IMAGE_SZ, IMAGE_SZ, 3],
name='DG_X')
# Create mask
mask = util.load_mask('mask.png')
# Load Places365 data
data = np.load('places/places_128.npz')
imgs = data[
'imgs_train'] # Originally from http://data.csail.mit.edu/places/places365/val_256.tar
imgs_p = util.preprocess_images_inpainting(imgs, mask=mask)
test_imgs = data['imgs_test']
test_imgs_p = util.preprocess_images_inpainting(test_imgs, mask=mask)
test_img = test_imgs[:N_TEST]
test_img_p = test_imgs_p[:N_TEST]
train_img = imgs[4, np.newaxis]
train_img_p = imgs_p[4, np.newaxis]
import matplotlib.pyplot as plt
from util import load_mask
from util import generate_masked_frame
import argparse
# python3.6 generate_image.py -device "iPhone 11 Pro.png" -screen "frame.png"
parser = argparse.ArgumentParser()
parser.add_argument('-device', help='device background')
parser.add_argument('-screen', help='screenshot')
args = parser.parse_args()
# load device background image and convert to hsv
background = cv2.cvtColor(cv2.imread(args.device), cv2.COLOR_BGR2RGB)
background_hsv = cv2.cvtColor(background, cv2.COLOR_RGB2HSV)
# generate mask
mask, mask_rect = load_mask(background_hsv)
x, y, w, h = mask_rect
# apply mask to screenshot
frame = cv2.imread(args.screen)
masked_frame, final_image = generate_masked_frame(background, mask, mask_rect,
frame)
f, ax = plt.subplots(1, 4, figsize=(20, 10))
ax[0].imshow(background)
ax[1].imshow(mask)
ax[2].imshow(masked_frame)
ax[3].imshow(final_image)
f.savefig("fig.png")
cv2.imwrite("out.png", cv2.cvtColor(final_image, cv2.COLOR_RGB2BGR))