def evaluate(self, test_filenames): total_mse = total_psnr = 0 if len(test_filenames) == 0: return 0, 0 for filename in test_filenames: mse = self.do_for_evaluate(filename, print_console=False) total_mse += mse total_psnr += util.get_psnr(mse, max_value=self.max_value) return total_mse / len(test_filenames), total_psnr / len(test_filenames)
def do_for_evaluate_with_output(self, file_path, output_directory, print_console=False):
filename, extension = os.path.splitext(file_path)
output_directory += "/" + self.name + "/"
util.make_dir(output_directory)
true_image = util.set_image_alignment(util.load_image(file_path, print_console=False), self.scale)
if true_image.shape[2] == 3 and self.channels == 1:
# for color images
input_y_image = loader.build_input_image(true_image, channels=self.channels, scale=self.scale,
alignment=self.scale, convert_ycbcr=True)
input_bicubic_y_image = util.resize_image_by_pil(input_y_image, self.scale,
resampling_method=self.resampling_method)
true_ycbcr_image = util.convert_rgb_to_ycbcr(true_image)
output_y_image = self.do(input_y_image, input_bicubic_y_image)
mse = util.compute_mse(true_ycbcr_image[:, :, 0:1], output_y_image,
border_size=self.psnr_calc_border_size)
loss_image = util.get_loss_image(true_ycbcr_image[:, :, 0:1], output_y_image,
border_size=self.psnr_calc_border_size)
output_color_image = util.convert_y_and_cbcr_to_rgb(output_y_image, true_ycbcr_image[:, :, 1:3])
util.save_image(output_directory + file_path, true_image)
util.save_image(output_directory + filename + "_input" + extension, input_y_image)
util.save_image(output_directory + filename + "_input_bicubic" + extension, input_bicubic_y_image)
util.save_image(output_directory + filename + "_true_y" + extension, true_ycbcr_image[:, :, 0:1])
util.save_image(output_directory + filename + "_result" + extension, output_y_image)
util.save_image(output_directory + filename + "_result_c" + extension, output_color_image)
util.save_image(output_directory + filename + "_loss" + extension, loss_image)
elif true_image.shape[2] == 1 and self.channels == 1:
# for monochrome images
input_image = loader.build_input_image(true_image, channels=self.channels, scale=self.scale,
alignment=self.scale)
input_bicubic_y_image = util.resize_image_by_pil(input_image, self.scale,
resampling_method=self.resampling_method)
output_image = self.do(input_image, input_bicubic_y_image)
mse = util.compute_mse(true_image, output_image, border_size=self.psnr_calc_border_size)
util.save_image(output_directory + file_path, true_image)
util.save_image(output_directory + filename + "_result" + extension, output_image)
else:
mse = 0
if print_console:
print("[%s] MSE:%f, PSNR:%f" % (filename, mse, util.get_psnr(mse)))
return mse
def main(not_parsed_args):
if len(not_parsed_args) > 1:
print("Unknown args:%s" % not_parsed_args)
exit()
model = DCSCN.SuperResolution(FLAGS, model_name=FLAGS.model_name)
model.load_datasets("training", FLAGS.data_dir + "/" + FLAGS.dataset,
FLAGS.batch_dir + "/" + FLAGS.dataset,
FLAGS.batch_image_size, FLAGS.stride_size)
model.load_datasets("test", FLAGS.data_dir + "/" + FLAGS.test_dataset,
FLAGS.batch_dir + "/" + FLAGS.test_dataset,
FLAGS.batch_image_size, FLAGS.stride_size)
model.build_graph()
model.build_optimizer()
model.build_summary_saver()
logging.info("\n" + str(sys.argv))
logging.info("Test Data:" + FLAGS.test_dataset + " Training Data:" +
FLAGS.dataset)
final_mse = final_psnr = 0
test_filenames = util.get_files_in_directory(FLAGS.data_dir + "/" +
FLAGS.test_dataset)
for i in range(FLAGS.tests):
train(model, FLAGS, i)
total_psnr = total_mse = 0
for filename in test_filenames:
mse = model.do_for_evaluate(filename,
FLAGS.output_dir,
output=i is (FLAGS.tests - 1))
total_mse += mse
total_psnr += util.get_psnr(mse, max_value=FLAGS.max_value)
logging.info("\nTrial(%d) %s" % (i, util.get_now_date()))
model.print_steps_completed(output_to_logging=True)
logging.info("MSE:%f, PSNR:%f\n" % (total_mse / len(test_filenames),
total_psnr / len(test_filenames)))
final_mse += total_mse
final_psnr += total_psnr
logging.info("=== summary [%d] %s [%s] ===" %
(FLAGS.tests, model.name, util.get_now_date()))
util.print_num_of_total_parameters(output_to_logging=True)
n = len(test_filenames) * FLAGS.tests
logging.info("\n=== Average [%s] MSE:%f, PSNR:%f ===" %
(FLAGS.test_dataset, final_mse / n, final_psnr / n))
def test(model, test_data):
test_filenames = util.get_files_in_directory(FLAGS.data_dir + "/" +
test_data)
total_psnr = total_mse = 0
for filename in test_filenames:
mse = model.do_for_evaluate_with_output(
filename, output_directory=FLAGS.output_dir, print_console=False)
total_mse += mse
total_psnr += util.get_psnr(mse)
logging.info("\n=== [%s] MSE:%f, PSNR:%f ===" %
(test_data, total_mse / len(test_filenames),
total_psnr / len(test_filenames)))
def train_batch(self):
_, mse = self.sess.run(
[self.training_optimizer, self.mse],
feed_dict={
self.x: self.batch_input,
self.x2: self.batch_input_quad,
self.y: self.batch_true_quad,
self.lr_input: self.lr,
self.dropout_input: self.dropout
})
self.training_psnr_sum += util.get_psnr(mse, max_value=self.max_value)
self.training_step += 1
self.step += 1
def test(model, test_data):
test_filenames = util.get_files_in_directory(FLAGS.data_dir + "/" +
test_data)
total_psnr = total_mse = 0
for filename in test_filenames:
mse = model.do_for_evaluate(filename,
output_directory=FLAGS.output_dir,
output=FLAGS.save_results)
total_mse += mse
total_psnr += util.get_psnr(mse, max_value=FLAGS.max_value)
logging.info("\n=== Average [%s] MSE:%f, PSNR:%f ===" %
(test_data, total_mse / len(test_filenames),
total_psnr / len(test_filenames)))
def train_batch(self):
feed_dict = {self.x: self.batch_input, self.x2: self.batch_input_bicubic, self.y: self.batch_true,
self.lr_input: self.lr, self.dropout: self.dropout_rate, self.is_training: 1}
if self.use_l1_loss:
_, loss = self.sess.run([self.training_optimizer, self.image_loss], feed_dict=feed_dict)
self.training_loss_sum += loss
else:
_, mse = self.sess.run([self.training_optimizer, self.mse], feed_dict=feed_dict)
self.training_loss_sum += mse
self.training_psnr_sum += util.get_psnr(mse, max_value=self.max_value)
self.training_step += 1
self.step += 1
def do_for_evaluate(self, file_path, output_directory="output", output=True, print_console=True):
filename, extension = os.path.splitext(file_path)
output_directory += "/"
true_image = util.set_image_alignment(util.load_image(file_path), self.scale)
if true_image.shape[2] == 3 and self.channels == 1:
input_y_image = loader.build_input_image(true_image, channels=self.channels, scale=self.scale,
alignment=self.scale, convert_ycbcr=True, jpeg_mode=self.jpeg_mode)
# for color images
if output:
input_bicubic_y_image = util.resize_image_by_pil(input_y_image, self.scale)
true_ycbcr_image = util.convert_rgb_to_ycbcr(true_image, jpeg_mode=self.jpeg_mode)
output_y_image = self.do(input_y_image, input_bicubic_y_image)
mse = util.compute_mse(true_ycbcr_image[:, :, 0:1], output_y_image, border_size=self.scale)
loss_image = util.get_loss_image(true_ycbcr_image[:, :, 0:1], output_y_image, border_size=self.scale)
output_color_image = util.convert_y_and_cbcr_to_rgb(output_y_image, true_ycbcr_image[:, :, 1:3],
jpeg_mode=self.jpeg_mode)
util.save_image(output_directory + file_path, true_image)
util.save_image(output_directory + filename + "_input" + extension, input_y_image)
util.save_image(output_directory + filename + "_input_bicubic" + extension, input_bicubic_y_image)
util.save_image(output_directory + filename + "_true_y" + extension, true_ycbcr_image[:, :, 0:1])
util.save_image(output_directory + filename + "_result" + extension, output_y_image)
util.save_image(output_directory + filename + "_result_c" + extension, output_color_image)
util.save_image(output_directory + filename + "_loss" + extension, loss_image)
else:
true_y_image = util.convert_rgb_to_y(true_image, jpeg_mode=self.jpeg_mode)
output_y_image = self.do(input_y_image)
mse = util.compute_mse(true_y_image, output_y_image, border_size=self.scale)
elif true_image.shape[2] == 1 and self.channels == 1:
# for monochrome images
input_image = loader.build_input_image(true_image, channels=self.channels, scale=self.scale, alignment=self.scale)
output_image = self.do(input_image)
mse = util.compute_mse(true_image, output_image, border_size=self.scale)
if output:
util.save_image(output_directory + file_path, true_image)
util.save_image(output_directory + filename + "_result" + extension, output_image)
if print_console:
print("MSE:%f PSNR:%f" % (mse, util.get_psnr(mse)))
return mse
def do_for_evaluate(self, file_path, print_console=False):
true_image = util.set_image_alignment(
util.load_image(file_path, print_console=False), self.scale)
if true_image.shape[2] == 3 and self.channels == 1:
# for color images
input_y_image = loader.build_input_image(true_image,
channels=self.channels,
scale=self.scale,
alignment=self.scale,
convert_ycbcr=True)
true_y_image = util.convert_rgb_to_y(true_image)
input_bicubic_y_image = util.resize_image_by_pil(
input_y_image,
self.scale,
resampling_method=self.resampling_method)
output_y_image = self.do(input_y_image, input_bicubic_y_image)
mse = util.compute_mse(true_y_image,
output_y_image,
border_size=self.psnr_calc_border_size)
elif true_image.shape[2] == 1 and self.channels == 1:
# for monochrome images
input_image = loader.build_input_image(true_image,
channels=self.channels,
scale=self.scale,
alignment=self.scale)
input_bicubic_y_image = util.resize_image_by_pil(
input_image,
self.scale,
resampling_method=self.resampling_method)
output_image = self.do(input_image, input_bicubic_y_image)
mse = util.compute_mse(true_image,
output_image,
border_size=self.psnr_calc_border_size)
else:
mse = 0
if print_console:
print("MSE:%f, PSNR:%f" % (mse, util.get_psnr(mse)))
return mse
def main(not_parsed_args):
if len(not_parsed_args) > 1:
print("Unknown args:%s" % not_parsed_args)
exit()
model = DCSCN.SuperResolution(FLAGS, model_name=FLAGS.model_name)
if FLAGS.build_batch:
model.load_datasets(FLAGS.data_dir + "/" + FLAGS.dataset,
FLAGS.batch_dir + "/" + FLAGS.dataset,
FLAGS.batch_image_size, FLAGS.stride_size)
else:
model.load_dynamic_datasets(FLAGS.data_dir + "/" + FLAGS.dataset,
FLAGS.batch_image_size)
model.build_graph()
model.build_optimizer()
model.build_summary_saver()
logging.info("\n" + str(sys.argv))
logging.info("Test Data:" + FLAGS.test_dataset + " Training Data:" +
FLAGS.dataset)
util.print_num_of_total_parameters(output_to_logging=True)
total_psnr = total_mse = 0
for i in range(FLAGS.tests):
mse = train(model, FLAGS, i)
psnr = util.get_psnr(mse, max_value=FLAGS.max_value)
total_mse += mse
total_psnr += psnr
logging.info("\nTrial(%d) %s" % (i, util.get_now_date()))
model.print_steps_completed(output_to_logging=True)
logging.info("MSE:%f, PSNR:%f\n" % (mse, psnr))
if FLAGS.tests > 1:
logging.info("\n=== Final Average [%s] MSE:%f, PSNR:%f ===" %
(FLAGS.test_dataset, total_mse / FLAGS.tests,
total_psnr / FLAGS.tests))
model.copy_log_to_archive("archive")
def train_batch(self, loader):
for batch in enumerate(loader.loader_train):
lr, hr = batch[1][0], batch[1][1]
feed_dict = {
self.x: lr.numpy(),
self.y: hr.numpy(),
self.lr_input: self.lr,
self.is_training: 1
}
_, mse = self.sess.run([self.training_optimizer, self.mae],
feed_dict=feed_dict)
self.training_mse_sum += mse
self.training_psnr_sum += util.get_psnr(mse,
max_value=self.max_value)
self.training_step += 1
self.step += 1
def update_epoch_and_lr(self, mse):
lr_updated = False
if self.min_validation_mse < 0 or self.min_validation_mse > mse:
# update new mse
self.min_validation_epoch = self.epochs_completed
self.min_validation_mse = mse
else:
if self.epochs_completed > self.min_validation_epoch + self.lr_decay_epoch:
# set new learning rate
self.min_validation_epoch = self.epochs_completed
self.lr *= self.lr_decay
lr_updated = True
psnr = util.get_psnr(mse, max_value=self.max_value)
self.csv_epochs.append(self.epochs_completed)
self.csv_psnr.append(psnr)
self.csv_training_psnr.append(self.training_psnr_sum / self.training_step)
return lr_updated
def test(model, test_data):
test_filenames = util.get_files_in_directory(FLAGS.data_dir + "/" +
test_data)
total_psnr = total_time = total_mse = 0
i = 0
for filename in test_filenames:
mse, time1 = model.do_for_evaluate(filename,
output_directory=FLAGS.output_dir,
output=FLAGS.save_results)
total_mse += mse
if i != 0:
total_time += time1
total_psnr += util.get_psnr(mse, max_value=FLAGS.max_value)
i = i + 1
logging.info("\n=== Average [%s] MSE:%f, PSNR:%f ===" %
(test_data, total_mse / len(test_filenames),
total_psnr / len(test_filenames)))
ave_time = total_time / (len(test_filenames) - 1)
#print ("end_t",end_t,"start_t",start_t)
print("total_time: %4.4f ave_time: %4.4f " % (total_time, ave_time))
def evaluate_test_batch(self,
save_meta_data=False,
trial=0,
log_profile=True):
save_meta_data = save_meta_data and self.save_meta_data and (trial
== 0)
feed_dict = {
self.x: self.test.input.images,
self.x2: self.test.input.hr_images,
self.y: self.test.true.hr_images,
self.dropout: 1.0,
self.is_training: 0
}
if log_profile and (self.save_loss or self.save_weights
or save_meta_data):
if save_meta_data:
# profiler = tf.profiler.Profile(self.sess.graph)
run_metadata = tf.RunMetadata()
run_options = tf.RunOptions(
trace_level=tf.RunOptions.FULL_TRACE)
summary_str, mse = self.sess.run([self.summary_op, self.mse],
feed_dict=feed_dict,
options=run_options,
run_metadata=run_metadata)
self.test_writer.add_run_metadata(
run_metadata, "step%d" % self.epochs_completed)
filename = self.checkpoint_dir + "/" + self.name + "_metadata.txt"
with open(filename, "w") as out:
out.write(str(run_metadata))
# filename = self.checkpoint_dir + "/" + self.name + "_memory.txt"
# tf.profiler.write_op_log(
# tf.get_default_graph(),
# log_dir=self.checkpoint_dir,
# #op_log=op_log,
# run_meta=run_metadata)
tf.contrib.tfprof.model_analyzer.print_model_analysis(
tf.get_default_graph(),
run_meta=run_metadata,
tfprof_options=tf.contrib.tfprof.model_analyzer.
PRINT_ALL_TIMING_MEMORY)
else:
summary_str, mse = self.sess.run([self.summary_op, self.mse],
feed_dict=feed_dict)
self.train_writer.add_summary(summary_str, self.epochs_completed)
util.log_scalar_value(self.train_writer, 'training_PSNR',
self.training_psnr_sum / self.training_step,
self.epochs_completed)
util.log_scalar_value(self.train_writer, 'LR', self.lr,
self.epochs_completed)
self.train_writer.flush()
util.log_scalar_value(self.test_writer, 'PSNR', util.get_psnr(mse),
self.epochs_completed)
self.test_writer.flush()
else:
mse = self.sess.run(self.mse, feed_dict=feed_dict)
return mse