def main(_):
data_dir = FLAGS.data_dir
label_map_dict = dataset_utils.parse_label_map_to_dict(
FLAGS.label_map_path)
logging.info('Reading from Pet dataset.')
image_dir = os.path.join(data_dir, 'images')
annotations_dir = os.path.join(data_dir, 'annotations')
examples_path = os.path.join(annotations_dir, 'trainval.txt')
examples_list = dataset_utils.read_examples_list(examples_path)
# Test images are not included in the downloaded data set, so we shall perform our own split.
random.seed(42)
random.shuffle(examples_list)
num_examples = len(examples_list)
num_train = int(0.7 * num_examples)
train_examples = examples_list[:num_train]
val_examples = examples_list[num_train:]
logging.info('%d training and %d validation examples.',
len(train_examples), len(val_examples))
train_output_path = os.path.join(FLAGS.output_dir, 'pet_train.txt')
val_output_path = os.path.join(FLAGS.output_dir, 'pet_val.txt')
def main():
if args.set not in sets:
raise ValueError('set must be in : {}'.format(sets))
data_dir = os.path.join(args.data_dir, args.set)
for label in labels:
writer = tf.python_io.TFRecordWriter(os.path.join(args.output_path, args.set+'_'+label+'.tfrecord'))
data_info = dataset_utils.read_examples_list(os.path.join(data_dir, label, label_file))
for image_name, theta in data_info:
image_path = os.path.join(data_dir, label, folder, image_name)
tf_example = dict_to_tf_example(image_path, args.image_size, label, theta)
writer.write(tf_example.SerializeToString())
writer.close()
def main(_):
if FLAGS.set not in SETS:
raise ValueError('set must be in : {}'.format(SETS))
if FLAGS.year not in YEARS:
raise ValueError('year must be in : {}'.format(YEARS))
data_dir = FLAGS.data_dir
years = ['VOC2007', 'VOC2012']
if FLAGS.year != 'merged':
years = [FLAGS.year]
output_file = os.path.join(data_dir,
'{}_{}.txt'.format(FLAGS.set, FLAGS.year))
with open(output_file, 'w') as txt_file:
items = 0
for year in years:
annotations_dir = os.path.join(data_dir, year,
FLAGS.annotations_dir)
examples_path = os.path.join(data_dir, year, 'ImageSets', 'Main',
'aeroplane_' + FLAGS.set + '.txt')
examples_list = read_examples_list(examples_path)
for idx, example in enumerate(examples_list):
path = os.path.join(annotations_dir, example + '.xml')
with tf.gfile.GFile(path, 'r') as fid:
xml_str = fid.read()
xml = etree.fromstring(xml_str)
data = recursive_parse_xml_to_dict(xml)['annotation']
img_path = os.path.join(data_dir, year, 'JPEGImages',
data['filename'])
for obj in data['object']:
difficult = bool(int(obj['difficult']))
if FLAGS.ignore_difficult_instances and difficult:
continue
xmin = float(obj['bndbox']['xmin'])
ymin = float(obj['bndbox']['ymin'])
xmax = float(obj['bndbox']['xmax'])
ymax = float(obj['bndbox']['ymax'])
classes_text = obj['name']
# Write item to text file here
items += 1
instance = "{},{},{},{},{},{}\n".format(
img_path, xmin, ymin, xmax, ymax, classes_text)
txt_file.write(instance)
print('Total instances', items)
def main():
if args.set not in sets:
raise ValueError('set must be in : {}'.format(sets))
data_dir = os.path.join(args.data_dir, args.set)
hparams = create_domain_adapt_hparams()
images_t = tf.placeholder(dtype=tf.float32, shape=[None, 224, 224, 3])
images_t = images_t - [123.68, 116.779, 103.939]
with slim.arg_scope(model_arg_scope()):
net, _ = model(inputs=images_t,
num_classes=num_classes,
is_training=False,
dropout_keep_prob=1.0,
reuse=tf.AUTO_REUSE,
scope=hparams.scope,
adapt_scope='target_adapt_layer',
adapt_dims=128)
min_index = tf.argmin(net, axis=1)
max_index = tf.argmax(net, axis=1)
saver = tf.train.Saver()
session_config = tf.ConfigProto(allow_soft_placement=True,
log_device_placement=False)
session_config.gpu_options.allow_growth = True
sess = tf.Session(config=session_config)
saver.restore(sess, tf.train.latest_checkpoint(args.checkpoint_dir))
print 'Successfully loading model: {}.'.format(
tf.train.latest_checkpoint(args.checkpoint_dir))
pos_writer = tf.python_io.TFRecordWriter(
os.path.join(args.output_path,
'pseudo_positive' + '_' + args.set + '.tfrecord'))
neg_writer = tf.python_io.TFRecordWriter(
os.path.join(args.output_path,
'pseudo_negative' + '_' + args.set + '.tfrecord'))
for label in labels:
data_info = dataset_utils.read_examples_list(
os.path.join(data_dir, label, label_file))
random.shuffle(data_info)
for image_name, theta in data_info:
image_path = os.path.join(data_dir, label, folder, image_name)
with tf.gfile.GFile(image_path, 'rb') as fid:
encoded_jpg = fid.read()
encoded_jpg_io = io.BytesIO(encoded_jpg)
image = PIL.Image.open(encoded_jpg_io)
if image.format != 'JPEG':
raise ValueError('Image format not JPEG')
image = image.resize((224, 224))
image_n = np.array(image)
image_n = np.expand_dims(image_n, axis=0).astype(np.float32)
scores, min_idx, max_idx = sess.run([net, min_index, max_index],
feed_dict={images_t: image_n})
if np.min(scores) < 0 and np.max(scores) > 0:
if np.min(scores) < -5.0:
tf_example = dict_to_tf_example(encoded_jpg, 'positive',
min_idx[0])
pos_writer.write(tf_example.SerializeToString())
if np.max(scores) > 5.0:
tf_example = dict_to_tf_example(encoded_jpg, 'negative',
max_idx[0])
neg_writer.write(tf_example.SerializeToString())
pos_writer.close()
neg_writer.close()