tf.summary.scalar('loss', loss)
tf.summary.scalar('acc', acc)
merged = tf.summary.merge_all()
def read_dcm(names, raw=False):
if raw:
img = sitk.ReadImage(names)
else:
names = sitk.ImageSeriesReader_GetGDCMSeriesFileNames(names)
img = sitk.ReadImage(names)
return sitk.GetArrayFromImage(img)
data = Data(path, BLOCK_SIZE)
if __name__ == '__main__':
with tf.Session(graph=g) as sess:
saver = tf.train.Saver()
tf.global_variables_initializer().run()
key = 0.0045 # 0.005
summary_writer = tf.summary.FileWriter('./summary', graph=sess.graph)
w = [0.1, 0.2, 0.3, 0.4]
ans3 = 1000
# cv2.imwrite('./prediction/test_.jpg', np.uint8(
# (pic[0, :, :, 0] < pic[0, :, :, 1])) * 255)
count = 0
iteration = 0
while iteration < 10000:
try:
try:
learning_rate=rates, momentum=0.2).minimize(loss=loss, global_step=g_steps)
with tf.variable_scope('evaluate'):
pre_img = tf.argmax(pre, -1)
ans = tf.equal(pre_img, Y)
acc = tf.reduce_mean(tf.cast(ans, tf.float32))
# VD = Metrics.VD(pre, Y)
# VOE = Metrics.VOE(pre, Y)
tf.summary.scalar('loss', loss)
tf.summary.scalar('acc', acc)
merged = tf.summary.merge_all()
pass
data = Data(path, BLOCK_SIZE)
if __name__ == '__main__':
with tf.Session(graph=g) as sess:
saver = tf.train.Saver()
# saver.restore(sess, './test_model_save_3/test.ckpt')
tf.global_variables_initializer().run()
key = 0.0045 # 0.005
# sess.run(tf.assign(g_steps, 0))
summary_writer = tf.summary.FileWriter('./summary', graph=sess.graph)
w = [0.1, 0.2, 0.3, 0.4]
ans3 = 1000
# cv2.imwrite('./prediction/test_.jpg', np.uint8(
# (pic[0, :, :, 0] < pic[0, :, :, 1])) * 255)
count = 0
iteration = 0
while iteration < 100000:
steps = 1000
g_steps = tf.Variable(0)
rates = tf.train.exponential_decay(0.2, g_steps, 200, 0.95, staircase=True)
# train = tf.train.GradientDescentOptimizer(rates).minimize(loss, global_step=g_steps)
train1 = tf.train.MomentumOptimizer(
learning_rate=rates, momentum=0.2).minimize(loss=loss_g, global_step=g_steps)
train2 = tf.train.MomentumOptimizer(
learning_rate=rates, momentum=0.2).minimize(loss=loss_d, global_step=g_steps, var_list=d_vars)
tf.summary.scalar('loss_d', loss_d)
tf.summary.scalar('loss_g', loss_g)
merged = tf.summary.merge_all()
data = Data(path, BLOCK_SIZE, stride)
if __name__ == '__main__':
with tf.Session(graph=g) as sess:
saver = tf.train.Saver()
saver.restore(sess, './test_model_save4000/test.ckpt')
key = 0.0045 # 0.005
summary_writer = tf.summary.FileWriter('./summary_1', graph=sess.graph)
w = [1, 2, 3, 4]
count = 0
iteration = 0
while iteration < 10000:
# try:
try:
x, y = data.next()
except Exception as e:
g_steps = tf.Variable(0)
rates = tf.train.exponential_decay(0.2, g_steps, 200, 0.95, staircase=True)
# train = tf.train.GradientDescentOptimizer(rates).minimize(loss, global_step=g_steps)
train1 = tf.train.MomentumOptimizer(
learning_rate=rates, momentum=0.2).minimize(loss=loss_g,
global_step=g_steps)
train2 = tf.train.MomentumOptimizer(
learning_rate=rates, momentum=0.2).minimize(loss=loss_d,
global_step=g_steps,
var_list=d_vars)
tf.summary.scalar('loss_d', loss_d)
tf.summary.scalar('loss_g', loss_g)
merged = tf.summary.merge_all()
data = Data(path, BLOCK_SIZE, stride)
if __name__ == '__main__':
with tf.Session(graph=g) as sess:
tf.global_variables_initializer().run()
x = np.zeros([1] + BLOCK_SIZE + [1])
# graph = tf.get_default_graph()
# tensor = graph.get_tensor_by_name('generator/block10/up/Relu:0')
ans = sess.run([tf.shape(monitor[0]),
tf.shape(supervise_stream[0])],
feed_dict={X: x})
# print(main_stream[10].name)
print(ans)
pass