def test_image_no_detection(self):
image = load_image(IMAGE_NO_DETECTION_PATH)
detected_image, classes = detect(self.tf_sess, image)
self.assertEqual((375, 500, 3), detected_image.shape)
self.assertListEqual([], sorted(classes))
def test_image_4k_detection(self):
image = load_image(IMAGE_4K_PATH)
detected_image, classes = detect(self.tf_sess, image)
self.assertEqual((2560, 3840, 3), detected_image.shape)
self.assertListEqual(['car', 'person'], sorted(classes))
def test_image_fullhd_detection(self):
image = load_image(IMAGE_FULL_HD_PATH)
detected_image, classes = detect(self.tf_sess, image)
self.assertEqual((1280, 1920, 3), detected_image.shape)
self.assertListEqual(['car', 'person'], sorted(classes))
def process_file_object(mc: Minio, tf_sess: tf.Session, bucket_name: str,
key: str, input_prefix: str, output_prefix: str):
file_name = os.path.basename(key)
print('Processing file: {}'.format(key))
try:
tmp_input_file_path = os.path.join('/tmp', file_name)
tmp_output_file_path = os.path.join('/tmp', 'output_' + file_name)
print('Downloading file: {}'.format(key))
try:
ret = mc.fget_object(bucket_name, key, tmp_input_file_path)
except NoSuchKey:
print('File not found, ignoring')
return
print('Loading image into memory')
try:
image_np = load_image(tmp_input_file_path)
finally:
os.remove(tmp_input_file_path)
print('Detecting objects in file: {}'.format(tmp_input_file_path))
image_np, classes = detect(tf_sess, image_np)
print('Detected classes: {}'.format(','.join(classes)))
metadata = {
**(ret.metadata or {}),
"x-amz-meta-classes": ','.join(classes),
}
try:
save_image(image_np, tmp_output_file_path)
output_key = key.replace(input_prefix, output_prefix, 1)
print('Uploading file to: {}'.format(output_key))
mc.fput_object(bucket_name,
output_key,
tmp_output_file_path,
metadata=metadata)
finally:
os.remove(tmp_output_file_path)
finally:
mc.remove_object(bucket_name, key)
print('Finished processing file')
def find_dates(tokens, postdate):
postdate = postdate.astimezone(get_time_zone())
all_dates = detect(tokens, postdate, TOLERANCE, tag_date, valid_state,
parse_date)
return process_dates(all_dates, tokens, postdate)
'--input_channel',
type=int,
default=1,
help='the number of input channel of Feature extractor')
parser.add_argument(
'--output_channel',
type=int,
default=512,
help='the number of output channel of Feature extractor')
parser.add_argument('--hidden_size',
type=int,
default=256,
help='the size of the LSTM hidden state')
opt = parser.parse_args()
images = detection.detect("test/milo.jpeg", "model/craft_mlt_25k.pth")
""" vocab / character number configuration """
if opt.sensitive:
opt.character = string.printable[:
-6] # same with ASTER setting (use 94 char).
cudnn.benchmark = True
cudnn.deterministic = True
opt.num_gpu = torch.cuda.device_count()
labels = ["indomi", "milo"]
recognition.demo(opt, 'model/TPS-ResNet-BiLSTM-Attn.pth', images, labels)
print("elapsed time : {}s".format(time.time() - t))