def on_epoch_end(self, batch, logs=None):
batches = len(self.validation_data)
total = batches * self.batch_size
real_total = 0
val_pred = np.zeros((total, self.num_classes))
val_true = np.zeros((total, self.num_classes))
for single_batch in range(batches):
val_x, val_y = next(self.validation_data)
row_num = val_x.shape[0]
real_total += total
val_pred[single_batch * self.batch_size: single_batch * self.batch_size + row_num] = \
prp_2_oh_array(np.asarray(self.model.predict(val_x)))
val_true[single_batch *
self.batch_size:single_batch * self.batch_size +
row_num] = val_y
val_pred = val_pred[:real_total, :]
val_true = val_true[:real_total, :]
warnings.filterwarnings('ignore', category=UndefinedMetricWarning)
precision, recall, f_score, support = precision_recall_fscore_support(
val_true, val_pred)
for p, r, f, s in zip(precision, recall, f_score, support):
print(
" - val_f1: %0.4f - val_pre: %0.4f - val_rec: %0.4f - ins %s" %
(f, p, r, s))
def on_epoch_end(self, batch, logs=None):
val_x = self.validation_data[0]
val_y = self.validation_data[1]
prd_y = prp_2_oh_array(np.asarray(self.model.predict(val_x)))
warnings.filterwarnings("ignore", category=UndefinedMetricWarning)
precision, recall, f_score, _ = precision_recall_fscore_support(
val_y, prd_y, average='macro')
print " — val_f1: % 0.4f — val_pre: % 0.4f — val_rec % 0.4f" % (f_score, precision, recall)
def on_epoch_end(self, batch, logs=None):
val_x = self.validation_data[0]
val_y = self.validation_data[1]
prd_y = prp_2_oh_array(np.asarray(self.model.predict(val_x)))
warnings.filterwarnings("ignore", category=UndefinedMetricWarning)
precision, recall, f_score, support = precision_recall_fscore_support(val_y, prd_y)
for p, r, f, s in zip(precision, recall, f_score, support):
print " — val_f1: % 0.4f — val_pre: % 0.4f — val_rec % 0.4f - ins %s" % (f, p, r, s)