def set_examples(self):
# List of (index root image, index target image, cross value)
# where index is index of image in the self.images_dict
self.examples = []
for source in self.data_sources:
print('Start set_examples from ' + source['code'])
progress_bar = Progbar(len(source['surface_match_data']))
for item in source['surface_match_data']:
root_frame_num = str(item[1]).zfill(4)
image_root_path = '\\' + source[
'code'] + '_images\\real\\scene-' + str(
item[0]) + '\\' + root_frame_num + '.jpg'
image_root_index = self.images_dict_flip[image_root_path]
target_frame_num = str(item[2]).zfill(4)
image_target_path = '\\' + source[
'code'] + '_images\\real\\scene-' + str(
item[0]) + '\\' + target_frame_num + '.jpg'
image_target_index = self.images_dict_flip[image_target_path]
self.examples.append(
(image_root_index, image_target_index, item[3]))
progress_bar.add(1)
def save_img(self):
progress_bar = Progbar(len(self.images_np_arr))
for i in range(len(self.images_np_arr)):
image = self.images_np_arr[i].astype('float32') / 255.
file = os.path.join(self.train_dir, 'images', str(i))
save_img(file + '.png', image)
if i % 100 == 0:
progress_bar.add(100)
def set_grouped_examples(self):
self.grouped_examples = [[] for i in range(GROUP_COUNT)]
progress_bar = Progbar(len(self.examples))
for example in self.examples:
example_value = example[2]
group = round(example_value * (GROUP_COUNT - 1))
self.grouped_examples[group].append(example)
progress_bar.add(1)
def calc_values_for_images(self):
print('Run calc_values_for_images()')
counter = 0
remain_images_to_calc = []
for image_file in self.image_files:
image_file_data = (image_file['scene'], image_file['root'],
image_file['frame'])
if self.is_image_calculated(image_file_data):
continue
remain_images_to_calc.append(image_file)
print('remain images to calc: ' + str(len(remain_images_to_calc)))
# Progress bar for remain items to calculate
progress_bar = Progbar(len(remain_images_to_calc))
for image_file in remain_images_to_calc:
img = Image.open(image_file['path'])
img.thumbnail((16, 16))
value = np.array(img).mean() / 255
self.data.append((
image_file['scene'],
image_file['root'],
image_file['frame'],
round(value, 4),
))
counter += 1
if counter % 5000 == 0:
progress_bar.add(5000)
if counter % 1000000 == 0 and counter > 0:
print('\nSave file partial with new records ' + str(counter))
self.save_data()
# Save data to json
print('\nSave file complete with new records ' + str(counter))
self.save_data()
int(indexes[index][0]),
int(indexes[index][1]),
delta
])
if delta > save_images_error_threshold and save_images:
r_val = 'r%.2f' % real_result
p_val = 'p%.2f' % predicted_result
root_name = str(batch) + '-' + str(index) + '-root_' + p_val + 'vs' + r_val + '.jpg'
target_name = str(batch) + '-' + str(index) + '-target_' + p_val + 'vs' + r_val + '.jpg'
save_img('bad_predictions/' + root_name, t_images_1[index])
save_img('bad_predictions/' + target_name, t_images_2[index])
if batch % 5 == 0 and batch > 0:
progress_bar.add(5)
if batch % 20 == 0 and batch > 0:
save_file(hard_indexes)
# Update weight complexity
batch_generator.load_example_weights()
for sample in samples:
batch_generator.update_weights(sample[0], sample[1], sample[2])
batch_generator.save_example_weights()
save_file(hard_indexes)