while input_var not in ['yes', 'no']:
input_var = input("We found model files. Do you want to load it and continue training [yes/no]?")
if input_var == 'yes':
load = True
if load:
saver.restore(sess, './weights/' + model_path + '.ckpt')
distances = []
test_labels = []
for i in range(1000):
if siamese.batch_size > 1:
x1_test, x2_test, sim_labels, x1_label, x2_label = dataset.get_batch(training=siamese.training,
optical_flow=siamese.optical_flow,
augment=False,
batch_size=siamese.batch_size,
seq_len=siamese.seq_len)
else:
if i % 2:
pair = dataset.get_positive_sequence_pair(training=siamese.training,
dense_optical_flow=siamese.optical_flow,
augment=False,
seq_len=siamese.seq_len)
else:
pair = dataset.get_negative_sequence_pair(training=siamese.training,
dense_optical_flow=siamese.optical_flow,
augment=False,
seq_len=siamese.seq_len)
)
if input_var == 'yes':
load = True
if load:
saver.restore(sess, './weights/' + model_path + '.ckpt')
losses_window = []
avg_loss = 0
for step in range(1000000):
if siamese.batch_size > 1:
batch_x1, batch_x2, batch_y, x1_label, x2_label = dataset.get_batch(
training=siamese.training,
optical_flow=siamese.optical_flow,
augment=True,
batch_size=siamese.batch_size,
seq_len=siamese.seq_len)
else:
if step % 2 == 0:
pair = dataset.get_positive_sequence_pair(
training=siamese.training,
dense_optical_flow=siamese.optical_flow,
augment=True,
seq_len=siamese.seq_len)
else:
pair = dataset.get_negative_sequence_pair(
training=siamese.training,
dense_optical_flow=siamese.optical_flow,
augment=True,