def test_video_audio_input(self):
params = exp_cfg.kinetics600(is_training=True)
params.feature_shape = (2, 224, 224, 3)
params.min_image_size = 224
params.output_audio = True
params.audio_feature = AUDIO_KEY
params.audio_feature_shape = (15, 256)
decoder = video_input.Decoder()
decoder.add_feature(params.audio_feature,
tf.io.VarLenFeature(dtype=tf.float32))
parser = video_input.Parser(params).parse_fn(params.is_training)
seq_example, label = fake_seq_example()
input_tensor = tf.constant(seq_example.SerializeToString())
decoded_tensors = decoder.decode(input_tensor)
output_tensor = parser(decoded_tensors)
features, label = output_tensor
image = features['image']
audio = features['audio']
self.assertAllEqual(image.shape, (2, 224, 224, 3))
self.assertAllEqual(label.shape, (600, ))
self.assertEqual(audio.shape, (15, 256))
def test_video_input(self):
params = exp_cfg.kinetics600(is_training=True)
params.feature_shape = (2, 224, 224, 3)
params.min_image_size = 224
decoder = video_input.Decoder()
parser = video_input.Parser(params).parse_fn(params.is_training)
# Create fake data.
random_image = np.random.randint(0, 256, size=(263, 320, 3), dtype=np.uint8)
random_image = Image.fromarray(random_image)
with io.BytesIO() as buffer:
random_image.save(buffer, format='JPEG')
raw_image_bytes = buffer.getvalue()
seq_example = tf.train.SequenceExample()
seq_example.feature_lists.feature_list.get_or_create(
video_input.IMAGE_KEY).feature.add().bytes_list.value[:] = [
raw_image_bytes
]
seq_example.feature_lists.feature_list.get_or_create(
video_input.IMAGE_KEY).feature.add().bytes_list.value[:] = [
raw_image_bytes
]
seq_example.context.feature[video_input.LABEL_KEY].int64_list.value[:] = [
42
]
input_tensor = tf.constant(seq_example.SerializeToString())
decoded_tensors = decoder.decode(input_tensor)
output_tensor = parser(decoded_tensors)
image_features, label = output_tensor
image = image_features['image']
self.assertAllEqual(image.shape, (2, 224, 224, 3))
self.assertAllEqual(label.shape, (600,))
def build_inputs(self, params: exp_cfg.DataConfig, input_context=None):
"""Builds classification input."""
parser = video_input.Parser(input_params=params)
postprocess_fn = video_input.PostBatchProcessor(params)
reader = input_reader.InputReader(
params,
dataset_fn=self._get_dataset_fn(params),
decoder_fn=self._get_decoder_fn(params),
parser_fn=parser.parse_fn(params.is_training),
postprocess_fn=postprocess_fn)
dataset = reader.read(input_context=input_context)
return dataset
def test_video_input(self):
params = exp_cfg.kinetics600(is_training=True)
params.feature_shape = (2, 224, 224, 3)
params.min_image_size = 224
decoder = video_input.Decoder()
parser = video_input.Parser(params).parse_fn(params.is_training)
seq_example, label = fake_seq_example()
input_tensor = tf.constant(seq_example.SerializeToString())
decoded_tensors = decoder.decode(input_tensor)
output_tensor = parser(decoded_tensors)
image_features, label = output_tensor
image = image_features['image']
self.assertAllEqual(image.shape, (2, 224, 224, 3))
self.assertAllEqual(label.shape, (600, ))
def build_inputs(
self,
params: exp_cfg.DataConfig,
input_context: Optional[tf.distribute.InputContext] = None):
"""Builds classification input."""
parser = video_input.Parser(input_params=params,
image_key=params.image_field_key,
label_key=params.label_field_key)
postprocess_fn = video_input.PostBatchProcessor(params)
reader = input_reader_factory.input_reader_generator(
params,
dataset_fn=self._get_dataset_fn(params),
decoder_fn=self._get_decoder_fn(params),
parser_fn=parser.parse_fn(params.is_training),
postprocess_fn=postprocess_fn)
dataset = reader.read(input_context=input_context)
return dataset
