"%s/%04d/predictions/final_%06d.jpg" % (model, epoch, ind),
np.uint8(
np.maximum(np.minimum(output[0] * 255.0, 255.0), 0.0)))
sic.imsave(
"%s/%04d/predictions/predictions_%06d.jpg" %
(model, epoch, ind),
np.uint8(
np.maximum(np.minimum(C0_imall[0] * 255.0, 255.0),
0.0)))
saver.save(sess, "%s/model.ckpt" % model)
saver.save(sess, "%s/%04d/model.ckpt" % (model, epoch))
# Inference
else:
test_low = utils.get_names(test_dir)
numtest = len(test_low)
print(test_low[0])
out_folder = test_dir.split('/')[-1]
outputs = [None] * 4
# for ind in range(numtest):
for ind in range(10):
input_image_src, input_image_target, input_flow_forward_src, input_flow_backward_src = prepare_input_w_flow(
test_low[ind], num_frames=num_frame, gray=True)
if input_image_src is None or input_image_target is None or input_flow_forward_src is None:
print("Not able to read the images/flows.")
continue
st = time.time()
C0_imall, C1_imall, C0_im, C1_im = sess.run(
[objDict["prediction_0"], objDict["prediction_1"], C0, C1],
feed_dict={
g1 = VCN(utils.build(tf.tile(input_i[:, :, :, 1:2], [1, 1, 1, 3])),
reuse=True)
saver = tf.compat.v1.train.Saver(max_to_keep=1000)
# +
sess.run([tf.compat.v1.global_variables_initializer()])
var_restore = [v for v in tf.compat.v1.trainable_variables()]
saver_restore = tf.compat.v1.train.Saver(var_restore)
ckpt = tf.train.get_checkpoint_state(model)
saver_restore.restore(sess, ckpt.model_checkpoint_path)
# -
if not len(test_img):
img_names = utils.get_names(test_dir)
ind = 0
for img_name in img_names:
im = np.float32(scipy.misc.imread(img_name, 'L')) / 255.0
h = im.shape[0] // 32 * 32
w = im.shape[1] // 32 * 32
im = im[np.newaxis, :h, :w, np.newaxis]
st = time.time()
output = sess.run(
g0, feed_dict={input_i: np.concatenate((im, im), axis=3)})
print("test time for %s --> %.3f" % (ind, time.time() - st))
folder = test_dir.split('/')[-1]
if not os.path.isdir("%s/%s" % (model, folder)):
for idx in range(5):
os.makedirs("%s/%s/result%d" % (model, folder, idx))
out_all = np.concatenate(