def test_procedure(self, test_data, distribution_op, inputX, inputY, mode):
confusion_matrics = np.zeros([self.num_class, self.num_class], dtype="int")
tst_batch_num = int(np.ceil(test_data[0].shape[0] / self.bs))
for step in range(tst_batch_num):
_testImg = test_data[0][step * self.bs:step * self.bs + self.bs]
_testLab = test_data[1][step * self.bs:step * self.bs + self.bs]
matrix_row, matrix_col = self.sess.run(distribution_op, feed_dict={inputX: _testImg,
inputY: _testLab,
self.is_training: False,
self.keep_rate: 0.5})
for m, n in zip(matrix_row, matrix_col):
confusion_matrics[m][n] += 1
test_accuracy = float(np.sum([confusion_matrics[q][q] for q in range(self.num_class)])) / float(
np.sum(confusion_matrics))
detail_test_accuracy = [confusion_matrics[i][i] / np.sum(confusion_matrics[i]) for i in
range(self.num_class)]
log0 = "Mode: " + mode
log1 = "Test Accuracy : %g" % test_accuracy
log2 = np.array(confusion_matrics.tolist())
log3 = ''
for j in range(self.num_class):
log3 += 'category %s test accuracy : %g\n' % (da_utils.pulmonary_category[j], detail_test_accuracy[j])
log3 = log3[:-1]
log4 = 'F_Value : %g\n' % self.f_value(confusion_matrics)
da_utils.save2file(log0, self.ckptDir, self.model)
da_utils.save2file(log1, self.ckptDir, self.model)
da_utils.save2file(log2, self.ckptDir, self.model)
da_utils.save2file(log3, self.ckptDir, self.model)
da_utils.save2file(log4, self.ckptDir, self.model)
def train(self):
print(
'Start to run in mode [Domain Adaptation Across Source and Target Domain]'
)
self.sess.run(tf.global_variables_initializer())
self.preTrained_saver = tf.train.Saver(var_list=self.g2_preTrained_var)
self.preTrained_saver.restore(self.sess, self.preTrained_path)
print('Pre-trained model has been successfully restored !')
self.train_itr = len(self.source_training_data[0]) // self.bs
for e in range(1, self.eps + 1):
_src_tr_img, _src_tr_lab = DA_init.shuffle_data(
self.source_training_data[0], self.source_training_data[1])
_tar_tr_img = DA_init.shuffle_data_nolabel(
self.target_training_data)
source_training_acc = 0.0
source_training_loss = 0.0
g_loss = 0.0
d_loss = 0.0
for itr in range(self.train_itr):
_src_tr_img_batch, _src_tr_lab_batch = DA_init.next_batch(
_src_tr_img, _src_tr_lab, self.bs, itr)
_tar_tr_img_batch = DA_init.next_batch_nolabel(
_tar_tr_img, self.bs)
feed_dict = {
self.x_source: _src_tr_img_batch,
self.y_source: _src_tr_lab_batch,
self.x_target: _tar_tr_img_batch,
self.is_training: True,
self.keep_prob: self.kp
}
feed_dict_eval = {
self.x_source: _src_tr_img_batch,
self.y_source: _src_tr_lab_batch,
self.x_target: _tar_tr_img_batch,
self.is_training: False,
self.keep_prob: 1.0
}
if e < 100:
_ = self.sess.run(self.g_train_op_step1,
feed_dict=feed_dict)
_training_accuracy, _training_loss = self.sess.run(
[self.accuracy_source, self.loss_source],
feed_dict=feed_dict_eval)
source_training_acc += _training_accuracy
source_training_loss += _training_loss
elif e < 200:
_, _ = self.sess.run(
[self.g_train_op_step2, self.d_train_op_step1],
feed_dict=feed_dict)
_training_accuracy, _training_loss, _g_loss, _d_loss = self.sess.run(
[
self.accuracy_source, self.loss_source,
self.g_loss_step2, self.d_loss_step1
],
feed_dict=feed_dict_eval)
source_training_acc += _training_accuracy
source_training_loss += _training_loss
g_loss += _g_loss
d_loss += _d_loss
elif e < self.eps:
_, _ = self.sess.run(
[self.g_train_op_step3, self.d_train_op_step2],
feed_dict=feed_dict)
_training_accuracy, _training_loss, _g_loss, _d_loss = self.sess.run(
[
self.accuracy_source, self.loss_source,
self.g_loss_step3, self.d_loss_step2
],
feed_dict=feed_dict_eval)
source_training_acc += _training_accuracy
source_training_loss += _training_loss
g_loss += _g_loss
d_loss += _d_loss
summary = self.sess.run(self.merged, feed_dict=feed_dict_eval)
source_training_acc = float(source_training_acc / self.train_itr)
source_training_loss = float(source_training_loss / self.train_itr)
g_loss = float(g_loss / self.train_itr)
d_loss = float(d_loss / self.train_itr)
source_validation_acc, source_validation_loss = self.validation_procedure(
validation_data=self.source_validation_data,
distribution_op=self.distribution_source,
loss_op=self.loss_source,
inputX=self.x_source,
inputY=self.y_source)
log1 = "Epoch: [%d], Domain: Source, Training Accuracy: [%g], Validation Accuracy: [%g], " \
"Training Loss: [%g], Validation Loss: [%g], generator Loss: [%g], Discriminator Loss: [%g], " \
"Time: [%s]" % (
e, source_training_acc, source_validation_acc, source_training_loss, source_validation_loss,
g_loss, d_loss, time.ctime(time.time()))
self.plt_epoch.append(e)
self.plt_training_accuracy.append(source_training_acc)
self.plt_training_loss.append(source_training_loss)
self.plt_validation_accuracy.append(source_validation_acc)
self.plt_validation_loss.append(source_validation_loss)
self.plt_d_loss.append(d_loss)
self.plt_g_loss.append(g_loss)
da_utils.plotAccuracy(x=self.plt_epoch,
y1=self.plt_training_accuracy,
y2=self.plt_validation_accuracy,
figName=self.model,
line1Name='training',
line2Name='validation',
savePath=self.ckptDir)
da_utils.plotLoss(x=self.plt_epoch,
y1=self.plt_training_loss,
y2=self.plt_validation_loss,
figName=self.model,
line1Name='training',
line2Name='validation',
savePath=self.ckptDir)
da_utils.plotLoss(x=self.plt_epoch,
y1=self.plt_d_loss,
y2=self.plt_g_loss,
figName=self.model + '_GD_Loss',
line1Name='D_Loss',
line2Name='G_Loss',
savePath=self.ckptDir)
da_utils.save2file(log1, self.ckptDir, self.model)
self.writer.add_summary(summary, e)
self.saver.save(self.sess,
self.ckptDir + self.model + '-' + str(e))
self.test_procedure(self.source_test_data,
distribution_op=self.distribution_source,
inputX=self.x_source,
inputY=self.y_source,
mode='source')
self.test_procedure(self.target_test_data,
distribution_op=self.distribution_target,
inputX=self.x_target,
inputY=self.y_target,
mode='target')
def saveConfiguration(self):
da_utils.save2file('epoch : %d' % self.eps, self.ckptDir, self.model)
da_utils.save2file('restore epoch : %d' % self.res_eps, self.ckptDir,
self.model)
da_utils.save2file('model : %s' % self.model, self.ckptDir, self.model)
da_utils.save2file('learning rate : %g' % self.lr, self.ckptDir,
self.model)
da_utils.save2file('batch size : %d' % self.bs, self.ckptDir,
self.model)
da_utils.save2file('image height : %d' % self.img_h, self.ckptDir,
self.model)
da_utils.save2file('image width : %d' % self.img_w, self.ckptDir,
self.model)
da_utils.save2file('num class : %d' % self.num_class, self.ckptDir,
self.model)
da_utils.save2file('train phase : %s' % self.train_phase, self.ckptDir,
self.model)
def train(self):
print('Start to run in mode [Domain Adaptation Across Source and Target Domain]')
self.sess.run(tf.global_variables_initializer())
self.reloadSaver = tf.train.Saver(var_list=self.g_var)
self.reloadSaver.restore(self.sess, self.reloadPath)
print('Pre-trained G2 model has been successfully reloaded !')
self.itr_epoch = len(self.source_training_data[0]) // self.bs
source_training_acc = 0.0
source_training_loss = 0.0
for e in range(1, self.eps + 1):
for itr in range(self.itr_epoch):
feed_dict_train, feed_dict_eval = self.getBatchData()
_ = self.sess.run(self.d_train_op, feed_dict=feed_dict_train)
feed_dict_train, feed_dict_eval = self.getBatchData()
_ = self.sess.run(self.g_train_op, feed_dict=feed_dict_train)
_training_accuracy, _training_loss = self.sess.run([self.accuracy_source, self.supervised_loss],
feed_dict=feed_dict_eval)
source_training_acc += _training_accuracy
source_training_loss += _training_loss
summary = self.sess.run(self.merged, feed_dict=feed_dict_eval)
source_training_acc = float(source_training_acc / self.itr_epoch)
source_training_loss = float(source_training_loss / self.itr_epoch)
source_validation_acc, source_validation_loss = self.validation_procedure(
validation_data=self.source_validation_data, distribution_op=self.distribution_source,
loss_op=self.supervised_loss, inputX=self.x_source, inputY=self.y_source)
log1 = "Epoch: [%d], Domain: Source, Training Accuracy: [%g], Validation Accuracy: [%g], " \
"Training Loss: [%g], Validation Loss: [%g], Time: [%s]" % (
e, source_training_acc, source_validation_acc, source_training_loss, source_validation_loss,
time.ctime(time.time()))
self.plt_epoch.append(e)
self.plt_training_accuracy.append(source_training_acc)
self.plt_training_loss.append(source_training_loss)
self.plt_validation_accuracy.append(source_validation_acc)
self.plt_validation_loss.append(source_validation_loss)
da_utils.plotAccuracy(x=self.plt_epoch,
y1=self.plt_training_accuracy,
y2=self.plt_validation_accuracy,
figName=self.model,
line1Name='training',
line2Name='validation',
savePath=self.ckptDir)
da_utils.plotLoss(x=self.plt_epoch,
y1=self.plt_training_loss,
y2=self.plt_validation_loss,
figName=self.model,
line1Name='training',
line2Name='validation',
savePath=self.ckptDir)
da_utils.save2file(log1, self.ckptDir, self.model)
self.writer.add_summary(summary, e)
self.saver.save(self.sess, self.ckptDir + self.model + '-' + str(e))
self.test_procedure(self.source_test_data, distribution_op=self.distribution_source,
inputX=self.x_source,
inputY=self.y_source, mode='source')
self.test_procedure(self.target_test_data, distribution_op=self.distribution_target,
inputX=self.x_target,
inputY=self.y_target, mode='target')
source_training_acc = 0.0
source_training_loss = 0.0
