def _train_with_summary(self):
img_batch, label_batch, labels, _ = self.tr_ds.get_next_batch(
self.sess)
feed = {
self.model.inputs: img_batch,
self.model.labels: label_batch,
self.model.is_training: True
}
fetches = [
self.model.total_loss, self.model.ctc_loss,
self.model.regularization_loss, self.model.global_step,
self.model.lr, self.model.merged_summay, self.model.dense_decoded,
self.model.edit_distance, self.model.train_op
]
batch_cost, _, _, global_step, lr, summary, predicts, edit_distance, _ = self.sess.run(
fetches, feed)
self.train_writer.add_summary(summary, global_step)
predicts = [
self.converter.decode(p, CRNN.CTC_INVALID_INDEX) for p in predicts
]
accuracy, _ = infer.calculate_accuracy(predicts, labels)
tf_utils.add_scalar_summary(self.train_writer, "train_accuracy",
accuracy, global_step)
tf_utils.add_scalar_summary(self.train_writer, "train_edit_distance",
edit_distance, global_step)
return batch_cost, global_step, lr
def _do_val(self, dataset, epoch, step, name, sess, model, converter, train_writer, cfg, result_dir):
if dataset is None:
return None
accuracy, edit_distance = infer.validation(sess, model.feeds(), model.fetches(),
dataset, converter, result_dir, name, step)
tf_utils.add_scalar_summary(train_writer, "%s_accuracy" % name, accuracy, step)
tf_utils.add_scalar_summary(train_writer, "%s_edit_distance" % name, edit_distance, step)
print("epoch: %d/%d, %s accuracy = %.3f" % (epoch, cfg.epochs, name, accuracy))
return accuracy
def _train_with_summary(self, model, tr_ds, sess, train_writer, converter):
img_batch, label_batch, labels, *rest = tr_ds.get_next_batch(sess)
image_batch_shape = img_batch.shape
w = self.round_up(image_batch_shape[2] / 4)
char_num = [len(l) for l in labels]
pos_init = [[-1, -1]]
# print('image_batch:',img_batch)
feed = {
model.inputs: img_batch,
model.labels: label_batch,
model.bat_labels: label_batch[1],
model.len_labels: w,
model.char_num: char_num,
model.char_pos_init: pos_init,
model.is_training: True
}
fetches = [
model.total_loss, model.ctc_loss, model.regularization_loss,
model.global_step, model.lr, model.merged_summary,
model.dense_decoded, model.edit_distance, model.train_op,
model.min_k, model.max_k
]
batch_cost, _, _, global_step, lr, summary, predicts, edit_distance, _, min_k, max_k = sess.run(
fetches, feed)
train_writer.add_summary(summary, global_step)
if min_k:
for k, (i, v, p) in enumerate(zip(*max_k)):
print(
'最大距离差的第 {} 个字符：[{}], 距离差：[{:.05}], prob：[{:.05}]'.format(
k, converter.decode_maps[i], v, p))
for k, (i, v, p) in enumerate(zip(*min_k)):
print(
'最小距离差的第 {} 个字符：[{}], 距离差：[{:.05}], prob：[{:.05}]'.format(
k, converter.decode_maps[i], v, p))
print(batch_cost)
predicts = [
converter.decode(p, CRNN.CTC_INVALID_INDEX) for p in predicts
]
accuracy, _ = infer.calculate_accuracy(predicts, labels)
tf_utils.add_scalar_summary(train_writer, "train_accuracy", accuracy,
global_step)
tf_utils.add_scalar_summary(train_writer, "train_edit_distance",
edit_distance, global_step)
return batch_cost, global_step, lr
def _train_with_summary(self, model, tr_ds, sess, train_writer,converter):
img_batch, label_batch, labels, positions, _ = tr_ds.get_next_batch(sess)
image_batch_shape = img_batch.shape
w = self.round_up(image_batch_shape[2]/4)
positions_list = []
for position_str in positions:
list2 = [6940 for i in range(w)]
position_list = str(position_str, encoding = "utf8").split(',')
num_list_new = [int(x) for x in position_list]
list2[:len(num_list_new)] = num_list_new
positions_list.append(list2)
con_labels_batch = np.array(positions_list)
con_labels_batch = con_labels_batch.reshape((-1))
# print('image_batch:',img_batch)
feed = {model.inputs: img_batch,
model.labels: label_batch,
model.con_labels: con_labels_batch,
model.len_labels: w,
model.is_training: True}
fetches = [model.total_loss,
model.ctc_loss,
model.regularization_loss,
model.global_step,
model.lr,
model.merged_summay,
model.dense_decoded,
model.edit_distance,
model.train_op]
batch_cost,_, _, global_step, lr, summary, predicts, edit_distance, _ = sess.run(fetches, feed)
train_writer.add_summary(summary, global_step)
print(batch_cost)
predicts = [converter.decode(p, CRNN.CTC_INVALID_INDEX) for p in predicts]
accuracy, _ = infer.calculate_accuracy(predicts, labels)
tf_utils.add_scalar_summary(train_writer, "train_accuracy", accuracy, global_step)
tf_utils.add_scalar_summary(train_writer, "train_edit_distance", edit_distance, global_step)
return batch_cost, global_step, lr