def _parse_fn(example_serialized, is_training):
"""Helper function for parse_fn_train() and parse_fn_valid()
Each Example proto (TFRecord) contains the following fields:
image/height: 462
image/width: 581
image/colorspace: 'RGB'
image/channels: 3
image/class/label: 615
image/class/synset: 'n03623198'
image/class/text: 'knee pad'
image/format: 'JPEG'
image/filename: 'ILSVRC2012_val_00041207.JPEG'
image/encoded: <JPEG encoded string>
Args:
example_serialized: scalar Tensor tf.string containing a
serialized Example protocol buffer.
is_training: training (True) or validation (False).
Returns:
image_buffer: Tensor tf.string containing the contents of
a JPEG file.
label: Tensor tf.int32 containing the label.
text: Tensor tf.string containing the human-readable label.
"""
feature_map = {
'image/encoded':
tf.FixedLenFeature([], dtype=tf.string, default_value=''),
'image/class/label':
tf.FixedLenFeature([], dtype=tf.int64, default_value=-1),
'image/class/text':
tf.FixedLenFeature([], dtype=tf.string, default_value=''),
}
parsed = tf.parse_single_example(example_serialized, feature_map)
image = decode_jpeg(parsed['image/encoded'])
if config.DATA_AUGMENTATION:
image = preprocess_image(image, 299, 299, is_training=is_training)
else:
image = resize_and_rescale_image(image, 299, 299)
# image = 1 - image
##tf.print(image)
#arr = tf.keras.preprocessing.image.img_to_array(image)
#print(arr.shape)
#with tf.Session() as sess:
# tf.print(sess.run(image))
#exit()
# The label in the tfrecords is 1~1000 (0 not used).
# So I think the minus 1 (of class label) is needed below.
label = tf.one_hot(parsed['image/class/label'] - 1, 1000, dtype=tf.float32)
return (image, label)
def run():
try:
frame1_url = request.args.get('img1')
frame2_url = request.args.get('img2')
frame1_path = get_image_from_url(https_url=frame1_url, file_id="1")
frame2_path = get_image_from_url(https_url=frame2_url, file_id="2")
processed_frame1_path = preprocess_image(frame_path=frame1_path, file_id="1")
processed_frame2_path = preprocess_image(frame_path=frame2_path, file_id="2")
_, similarity = compare_two_images(frame1_path=processed_frame1_path, frame2_path=processed_frame2_path)
ss = 1 - math.exp(EXP_CONST * similarity)
data = {'score': ss}
response = json.dumps(data)
for path in glob.glob(os.path.join(CUR_DIR, 'utils', '*.jpg')):
os.remove(path)
return response
except Exception as e:
log_print(info_str=e)
data = {'score': 0}
response = json.dumps(data)
return response
def _parse_fn(example_serialized, is_training):
"""Helper function for parse_fn_train() and parse_fn_valid()
Each Example proto (TFRecord) contains the following fields:
image/height: 462
image/width: 581
image/colorspace: 'RGB'
image/channels: 3
image/class/label: 615
image/class/synset: 'n03623198'
image/class/text: 'knee pad'
image/format: 'JPEG'
image/filename: 'ILSVRC2012_val_00041207.JPEG'
image/encoded: <JPEG encoded string>
Args:
example_serialized: scalar Tensor tf.string containing a
serialized Example protocol buffer.
Returns:
image_buffer: Tensor tf.string containing the contents of
a JPEG file.
label: Tensor tf.int32 containing the label.
text: Tensor tf.string containing the human-readable label.
"""
feature_map = {
'image/encoded':
tf.FixedLenFeature([], dtype=tf.string, default_value=''),
'image/class/label':
tf.FixedLenFeature([], dtype=tf.int64, default_value=-1),
'image/class/text':
tf.FixedLenFeature([], dtype=tf.string, default_value=''),
}
parsed = tf.parse_single_example(example_serialized, feature_map)
image = decode_jpeg(parsed['image/encoded'])
if config.DATA_AUGMENTATION:
image = preprocess_image(image, 224, 224, is_training=is_training)
else:
image = resize_and_rescale_image(image, 224, 224)
label = tf.one_hot(parsed['image/class/label'], 1000, dtype=tf.float32)
return (image, label)