def test_print(self):
y_true = np.array([0, 1, 2, 0])
y_pred = y_true
# expected_confusion_matrix = np.array([[2, 0, 0], [0, 1, 0], [0, 0, 1]])
metrics.report_multiclass_metrics(y_true, y_pred)
y_pred = [[0, 0, 1], [1, 1, 0], [0, 1, 1]]
y_true = [[0, 1, 1], [1, 1, 0], [0, 1, 1]]
# expected confusion matix:
# Confusion matrix for label 0:
# [[2 0]
# [0 1]]
# Confusion matrix for label 1:
# [[0 0]
# [1 2]]
# Confusion matrix for label 2:
# [[1 0]
# [0 2]]
metrics.report_multilabel_confusion_matrix(y_true, y_pred)
# expected confusion matrix:
# Confusion matrix for label 0:
# [[2 0]
# [0 1]]
# Confusion matrix for label 1:
# [[3]]
# Confusion matrix for label 2:
# [[1 0]
# [0 2]]
metrics.report_multilabel_confusion_matrix(y_true, y_true)
def eval(self, feed_dict):
print("========= Evaluation =========")
_ = self.tf_session.run(self.dataset_initializer, feed_dict=feed_dict)
loss_value, confusion_matrix, class_scores_values, predictions_class_winners_values, labels_winners_values = \
self.tf_session.run(
[self._loss_op, self._confusion_matrix, self._class_scores, self._predictions_class_winners,
self._labels_winners])
print(f'Loss: {loss_value:.2f}')
metrics.report_multiclass_metrics(labels_winners_values, predictions_class_winners_values)
print("========= Evaluation Complete =========\n\n")
def train(self, feed_dict):
print("========= Training =========")
_ = self.tf_session.run(self.dataset_initializer, feed_dict=feed_dict)
for step in range(self._max_training_steps):
_, loss_value, confusion_matrix, class_scores_values, predictions_class_winners_values, \
labels_winners_values = self.tf_session.run(
[self._train_op, self._loss_op, self._confusion_matrix, self._class_scores,
self._predictions_class_winners, self._labels_winners])
summary_str = self.tf_session.run(self.summary, feed_dict=feed_dict)
if self._write_summary:
self.summary_writer.add_summary(summary_str, step)
self.summary_writer.flush()
if step % int(self._max_training_steps / 20) == 0:
print(f'\n-----')
print(f'Step {step}')
print(f'Loss: {loss_value:.2f}')
metrics.report_multiclass_metrics(labels_winners_values, predictions_class_winners_values)
print("========= Training Complete =========\n\n")
def test_print(self):
y_true = np.array([0, 1, 2, 0])
y_pred = y_true
# expected_confusion_matrix = np.array([[2, 0, 0], [0, 1, 0], [0, 0, 1]])
metrics.report_multiclass_metrics(y_true, y_pred)
def _report_metrics(self, labels, predictions):
return metrics.report_multiclass_metrics(labels, predictions)