# momentum=0.9,use_nesterov=True)
# train = opt_func.minimize(loss)
tvars = tf.trainable_variables()
grads, _ = tf.clip_by_global_norm(tf.gradients(loss, tvars), 1)
optimizer = opt_func.apply_gradients(zip(grads, tvars),
global_step=global_step)
# Summaries
# tensorboard --logdir=...
with tf.name_scope('summaries'):
tf.summary.scalar('loss', loss)
tf.summary.scalar('psnr', psnr(x_result, x_true, 'compute_psnr'))
tf.summary.image('x_result', x_result)
tf.summary.image('x_true', x_true)
tf.summary.image('squared_error', squared_error)
tf.summary.image('residual', residual)
merged_summary = tf.summary.merge_all()
test_summary_writer = tf.summary.FileWriter(adler.tensorflow.util.default_tensorboard_dir(name) + '/test',
sess.graph)
train_summary_writer = tf.summary.FileWriter(adler.tensorflow.util.default_tensorboard_dir(name) + '/train')
# Initialize all TF variables
sess.run(tf.global_variables_initializer())
# Add op to save and restore
[primal_values, dual_values],
feed_dict={
x_true: x_true_arr_validate,
y_rt: y_arr_validate,
is_training: False
})
import matplotlib.pyplot as plt
from skimage.measure import compare_ssim as ssim
from skimage.measure import compare_psnr as psnr
print(ssim(primal_values_result[-1][0, ..., 0], x_true_arr_validate[0, ...,
0]))
print(
psnr(primal_values_result[-1][0, ..., 0],
x_true_arr_validate[0, ..., 0],
dynamic_range=np.max(x_true_arr_validate) -
np.min(x_true_arr_validate)))
def normalized(val, sign=False):
if sign:
val = val * np.sign(np.mean(val))
return (val - np.mean(val)) / np.std(val)
path = name
for i in range(n_iter):
vals = primal_values_result[i]
space.element(vals[..., 0]).show(saveto='{}/x_{}'.format(path, i))
space.element(vals[..., 0]).show(clim=[0.8, 1.2],
saveto='{}/x_windowed_{}'.format(path, i))