tf-models-nightly 2.17.0.dev20240422__py2.py3-none-any.whl → 2.17.0.dev20240423__py2.py3-none-any.whl

official/recommendation/uplift/metrics/poisson_metrics.py

@@ -130,6 +130,7 @@ def _safe_x_minus_xlogx(x: tf.Tensor) -> tf.Tensor:
   return tf.where(tf.equal(x, 0.0), tf.zeros_like(x), values)
 
 
+@tf_keras.utils.register_keras_serializable(package="Uplift")
 class LogLossMeanBaseline(tf_keras.metrics.Metric):
   """Computes the (weighted) poisson log loss for a mean predictor."""
 
@@ -200,6 +201,7 @@ class LogLossMeanBaseline(tf_keras.metrics.Metric):
     return cls(**config)
 
 
+@tf_keras.utils.register_keras_serializable(package="Uplift")
 class LogLossMinimum(tf_keras.metrics.Metric):
   """Computes the minimum achievable (weighted) poisson log loss.
 
@@ -278,6 +280,7 @@ class LogLossMinimum(tf_keras.metrics.Metric):
     return cls(**config)
 
 
+@tf_keras.utils.register_keras_serializable(package="Uplift")
 class PseudoRSquared(tf_keras.metrics.Metric):
   """Computes the pseudo R-squared metric for poisson regression.
 

official/vision/configs/video_classification.py

@@ -59,6 +59,7 @@ class DataConfig(cfg.DataConfig):
   aug_max_aspect_ratio: float = 2.0
   aug_min_area_ratio: float = 0.49
   aug_max_area_ratio: float = 1.0
+  aug_random_rotation: bool = False
   aug_type: Optional[
       common.Augmentation] = None  # AutoAugment and RandAugment.
   mixup_and_cutmix: Optional[common.MixupAndCutmix] = None

official/vision/dataloaders/video_input.py

@@ -44,6 +44,7 @@ def process_image(image: tf.Tensor,
                   max_aspect_ratio: float = 2,
                   min_area_ratio: float = 0.49,
                   max_area_ratio: float = 1.0,
+                  random_rotation: bool = False,
                   augmenter: Optional[augment.ImageAugment] = None,
                   seed: Optional[int] = None,
                   input_image_format: Optional[str] = 'jpeg') -> tf.Tensor:
@@ -77,10 +78,11 @@ def process_image(image: tf.Tensor,
     max_aspect_ratio: The maximum aspect range for cropping.
     min_area_ratio: The minimum area range for cropping.
     max_area_ratio: The maximum area range for cropping.
+    random_rotation: Use uniform random rotation augmentation or not.
     augmenter: Image augmenter to distort each image.
     seed: A deterministic seed to use when sampling.
     input_image_format: The format of input image which could be jpeg, png or
-          none for unknown or mixed datasets.
+      none for unknown or mixed datasets.
 
   Returns:
     Processed frames. Tensor of shape
@@ -136,6 +138,8 @@ def process_image(image: tf.Tensor,
         (min_aspect_ratio, max_aspect_ratio),
         (min_area_ratio, max_area_ratio))
     image = preprocess_ops_3d.random_flip_left_right(image, seed)
+    if random_rotation:
+      image = preprocess_ops_3d.random_rotation(image, seed)
 
     if augmenter is not None:
       image = augmenter.distort(image)
@@ -303,6 +307,7 @@ class Parser(parser.Parser):
     self._min_area_ratio = input_params.aug_min_area_ratio
     self._max_area_ratio = input_params.aug_max_area_ratio
     self._input_image_format = input_params.input_image_format
+    self._random_rotation = input_params.aug_random_rotation
     if self._output_audio:
       self._audio_feature = input_params.audio_feature
       self._audio_shape = input_params.audio_feature_shape
@@ -329,6 +334,10 @@ class Parser(parser.Parser):
             'Augmentation policy {} not supported.'.format(aug_type.type))
     else:
       self._augmenter = None
+      if self._random_rotation:
+        logging.info('Using standard augmentation with rotation.')
+      else:
+        logging.info('Using standard augmentation without rotation.')
 
   def _parse_train_data(
       self, decoded_tensors: Dict[str, tf.Tensor]
@@ -350,6 +359,7 @@ class Parser(parser.Parser):
         max_aspect_ratio=self._max_aspect_ratio,
         min_area_ratio=self._min_area_ratio,
         max_area_ratio=self._max_area_ratio,
+        random_rotation=self._random_rotation,
         augmenter=self._augmenter,
         zero_centering_image=self._zero_centering_image,
         input_image_format=self._input_image_format)

official/vision/ops/preprocess_ops_3d.py

@@ -404,9 +404,33 @@ def random_flip_left_right(frames: tf.Tensor,
   return frames
 
 
-def normalize_image(frames: tf.Tensor,
-                    zero_centering_image: bool,
-                    dtype: tf.dtypes.DType = tf.float32) -> tf.Tensor:
+def random_rotation(frames: tf.Tensor, seed: Optional[int] = None) -> tf.Tensor:
+  """Randomly rotate all frames with 0, 90, 180, or 270 degrees.
+
+  Args:
+    frames: A Tensor of shape [timesteps, input_h, input_w, channels].
+    seed: A seed to use for the random sampling.
+
+  Returns:
+    A Tensor of shape [timesteps, output_h, output_w, channels] eventually
+    rotated at 0/90/180/270 degrees.
+  """
+  rotation_times = tf.random.uniform(
+      (), minval=0, maxval=4, dtype=tf.int32, seed=seed
+  )
+  frames = tf.cond(
+      tf.greater(rotation_times, 0),
+      true_fn=lambda: tf.image.rot90(frames, k=rotation_times),
+      false_fn=lambda: frames,
+  )
+  return frames
+
+
+def normalize_image(
+    frames: tf.Tensor,
+    zero_centering_image: bool,
+    dtype: tf.dtypes.DType = tf.float32,
+) -> tf.Tensor:
   """Normalizes images.
 
   Args:

official/vision/ops/preprocess_ops_3d_test.py

@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-
 import io
 import itertools
 import numpy as np
@@ -181,6 +180,25 @@ class ParserUtilsTest(tf.test.TestCase):
     self.assertTrue((flipped_frames == self._np_frames).numpy().all() or (
         flipped_frames == flipped).numpy().all())
 
+  def test_random_rotation(self):
+    rotated_frames = preprocess_ops_3d.random_rotation(self._frames)
+
+    rotated_once = np.rot90(self._np_frames[0, :, :, 0], 1)
+    rotated_twice = np.rot90(self._np_frames[0, :, :, 0], 2)
+    rotated_thrice = np.rot90(self._np_frames[0, :, :, 0], 3)
+    rotated_once = rotated_once[np.newaxis, :, :, np.newaxis]
+    rotated_twice = rotated_twice[np.newaxis, :, :, np.newaxis]
+    rotated_thrice = rotated_thrice[np.newaxis, :, :, np.newaxis]
+    rotated_once = np.broadcast_to(rotated_once, (6, 120, 90, 3))
+    rotated_twice = np.broadcast_to(rotated_twice, (6, 90, 120, 3))
+    rotated_thrice = np.broadcast_to(rotated_thrice, (6, 120, 90, 3))
+    self.assertTrue(
+        (rotated_frames == self._np_frames).numpy().all()
+        or (rotated_frames == rotated_once).numpy().all()
+        or (rotated_frames == rotated_twice).numpy().all()
+        or (rotated_frames == rotated_thrice).numpy().all()
+    )
+
   def test_normalize_image(self):
     normalized_images_1 = preprocess_ops_3d.normalize_image(
         self._frames, False, tf.float32)

{tf_models_nightly-2.17.0.dev20240422.dist-info → tf_models_nightly-2.17.0.dev20240423.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: tf-models-nightly
-Version: 2.17.0.dev20240422
+Version: 2.17.0.dev20240423
 Summary: TensorFlow Official Models
 Home-page: https://github.com/tensorflow/models
 Author: Google Inc.

{tf_models_nightly-2.17.0.dev20240422.dist-info → tf_models_nightly-2.17.0.dev20240423.dist-info}/RECORD

@@ -915,7 +915,7 @@ official/recommendation/uplift/metrics/label_variance_test.py,sha256=k0mdEU1WU53
 official/recommendation/uplift/metrics/loss_metric.py,sha256=gYZdnTsuL_2q1FZuPip-DaWxt_Q-02YYaePyMBVNx7w,7344
 official/recommendation/uplift/metrics/loss_metric_test.py,sha256=48rQG8bKFdy0xBFjoOLXKRUlYpCEyAzSmPOFoF7FX94,16021
 official/recommendation/uplift/metrics/metric_configs.py,sha256=Z-r79orE4EycQ5TJ7xdI5LhjOHT3wzChYyDxcxGqLXk,1670
-official/recommendation/uplift/metrics/poisson_metrics.py,sha256=ZltEopHvYJlrqjVr9X_NhNQOwWcv2O3rFqvKVVTC57I,14475
+official/recommendation/uplift/metrics/poisson_metrics.py,sha256=YV-5RgwOBihRoTgmbEYmnKKLPMxRhQjLGJ4jmDsGUVo,14661
 official/recommendation/uplift/metrics/poisson_metrics_test.py,sha256=zGo7Y7XJY4kvecoeAy7Jci_N3YZKMTKpj28j7ZtoTLc,16378
 official/recommendation/uplift/metrics/sliced_metric.py,sha256=uhvzudOWtMNKZ0avwGhX-37UELR9Cq9b4C0g8erBkXw,8688
 official/recommendation/uplift/metrics/sliced_metric_test.py,sha256=bhVGyI1tOkFkVOtruJo3p6XopDFyG1JW5qdZm9-RqeU,12248
@@ -969,7 +969,7 @@ official/vision/configs/retinanet.py,sha256=oCKinkh4IyPslmI1pakwi6dVziwjkZ2cIcpS
 official/vision/configs/retinanet_test.py,sha256=ffS3QufQMLF8FZhKNmi7Yr1RDTnIyZ1XKQ9agr2EyW8,1699
 official/vision/configs/semantic_segmentation.py,sha256=4ZAyLWKcFYReyrEWBc5b7wld3mMcuH0RcaRe_4J2RrA,30831
 official/vision/configs/semantic_segmentation_test.py,sha256=va-ZG6CtBKcs0NicZe6WmJvHxPxxih7nB0orNtrRiEA,1867
-official/vision/configs/video_classification.py,sha256=tf2XJhD_7c1Ned3eS93Sc0qrQ8U3M_zVJy09KI-9em8,14513
+official/vision/configs/video_classification.py,sha256=DDiBq9tCfk3x-FNprWvWc5hLaVAjGLTWgGvXCpElbw4,14549
 official/vision/configs/video_classification_test.py,sha256=I1HSamxRQ3-f4-YHIeUChnT5CtHCxFQdiL0zy6RRUXU,1879
 official/vision/data/__init__.py,sha256=7oiypy0N82PDw9aSdcJBLVoGTd_oRSUOdvuJhMv4leQ,609
 official/vision/data/create_coco_tf_record.py,sha256=UOvAPYONUAkoeWMtE18fVa6qfqzu_kulrwDkEpUXcjM,22742
@@ -1003,7 +1003,7 @@ official/vision/dataloaders/tfds_segmentation_decoders.py,sha256=-T8hrcEjaUivvtQ
 official/vision/dataloaders/tfexample_utils.py,sha256=qYUaT3V2zcEybJooImTxRT6MbCYzKAc_6iZPzIgBiyw,12698
 official/vision/dataloaders/utils.py,sha256=0E20HOYq-NeK479bmVmXQakD8M9GQn_lyWeyVXRdWq4,3113
 official/vision/dataloaders/utils_test.py,sha256=8gG1b0Wx8mjvAxEEnqSOBIXOGNGQM0zAKpX815umKu4,2567
-official/vision/dataloaders/video_input.py,sha256=sCLyuOPo7urUbXQSOWirlBoisBBWvJmZWENKAFKY5e8,17267
+official/vision/dataloaders/video_input.py,sha256=4TuPmV3W0c5vcNPeQNX8FdlCXqz3fGecss9vMi5WjJs,17756
 official/vision/dataloaders/video_input_test.py,sha256=fhHh5lft0KthdJ0_6n3kP3dZ6y8GLkji1Rlq4zP0jbU,7551
 official/vision/evaluation/__init__.py,sha256=7oiypy0N82PDw9aSdcJBLVoGTd_oRSUOdvuJhMv4leQ,609
 official/vision/evaluation/coco_evaluator.py,sha256=Nv6ElDBw2CSHFfogWpnAgCnIlMwmt5SWDbt-YeIvR3s,15601
@@ -1118,8 +1118,8 @@ official/vision/ops/mask_ops.py,sha256=cZLpIowzEA57bXPDbVXa6mktZVHvGSH-TQ1CxHjpQ
 official/vision/ops/mask_ops_test.py,sha256=D3xbbbleJd4HkpWOSDSEy6hNihsRBY93BqPF6JP-dJk,2835
 official/vision/ops/nms.py,sha256=bKYDAtyV5j6PG7g-RGF2ZccCI5V1xVvuajNblCy1TGs,8125
 official/vision/ops/preprocess_ops.py,sha256=tDW9apUocwSzGnKkgSrKGwqbcZpb6-WQ8nGnW1_ds20,42569
-official/vision/ops/preprocess_ops_3d.py,sha256=K2583ynA9Zt9pOBNoWSD8KtQR1fwRYBoylJ9NusIBtI,16110
-official/vision/ops/preprocess_ops_3d_test.py,sha256=1MmygNfRBnQYGszsrKBGqP_GrPlP4_UGuddCbkYcIms,8364
+official/vision/ops/preprocess_ops_3d.py,sha256=o0ivTBK2BwaiPp_NhVG_yBc10VUySxfE7eKQkL7RNaU,16762
+official/vision/ops/preprocess_ops_3d_test.py,sha256=LA3-Ue4FTjdsr5Kl_BnpAMNcfikWylMisD2GrBTOLzA,9309
 official/vision/ops/preprocess_ops_test.py,sha256=FY3EUvQIb82fYqYJPmzkE7pmkhXIQrd7JNLGFPB6SXI,17213
 official/vision/ops/sampling_ops.py,sha256=1jywCA_E4qDUFWsykCLUwZsWtQSR0QREXdJhvP5qCvc,16072
 official/vision/ops/spatial_transform_ops.py,sha256=PVEJGAn0ygtsrid84vD5GgV0jsjyWoNn14RBzreMxM4,38389
@@ -1206,9 +1206,9 @@ tensorflow_models/tensorflow_models_test.py,sha256=nc6A9K53OGqF25xN5St8EiWvdVbda
 tensorflow_models/nlp/__init__.py,sha256=4tA5Pf4qaFwT-fIFOpX7x7FHJpnyJT-5UgOeFYTyMlc,807
 tensorflow_models/uplift/__init__.py,sha256=mqfa55gweOdpKoaQyid4A_4u7xw__FcQeSIF0k_pYmI,999
 tensorflow_models/vision/__init__.py,sha256=zBorY_v5xva1uI-qxhZO3Qh-Dii-Suq6wEYh6hKHDfc,833
-tf_models_nightly-2.17.0.dev20240422.dist-info/AUTHORS,sha256=1dG3fXVu9jlo7bul8xuix5F5vOnczMk7_yWn4y70uw0,337
-tf_models_nightly-2.17.0.dev20240422.dist-info/LICENSE,sha256=WxeBS_DejPZQabxtfMOM_xn8qoZNJDQjrT7z2wG1I4U,11512
-tf_models_nightly-2.17.0.dev20240422.dist-info/METADATA,sha256=V1oJinOVeHP7XkD3Iqa_Q-xCEt5B4UXqF-TZnfYZU-4,1432
-tf_models_nightly-2.17.0.dev20240422.dist-info/WHEEL,sha256=kGT74LWyRUZrL4VgLh6_g12IeVl_9u9ZVhadrgXZUEY,110
-tf_models_nightly-2.17.0.dev20240422.dist-info/top_level.txt,sha256=gum2FfO5R4cvjl2-QtP-S1aNmsvIZaFFT6VFzU0f4-g,33
-tf_models_nightly-2.17.0.dev20240422.dist-info/RECORD,,
+tf_models_nightly-2.17.0.dev20240423.dist-info/AUTHORS,sha256=1dG3fXVu9jlo7bul8xuix5F5vOnczMk7_yWn4y70uw0,337
+tf_models_nightly-2.17.0.dev20240423.dist-info/LICENSE,sha256=WxeBS_DejPZQabxtfMOM_xn8qoZNJDQjrT7z2wG1I4U,11512
+tf_models_nightly-2.17.0.dev20240423.dist-info/METADATA,sha256=kTJSu84yICl-fJ2_SP7HYjLNi0CEFgUfn2IULGSkst0,1432
+tf_models_nightly-2.17.0.dev20240423.dist-info/WHEEL,sha256=kGT74LWyRUZrL4VgLh6_g12IeVl_9u9ZVhadrgXZUEY,110
+tf_models_nightly-2.17.0.dev20240423.dist-info/top_level.txt,sha256=gum2FfO5R4cvjl2-QtP-S1aNmsvIZaFFT6VFzU0f4-g,33
+tf_models_nightly-2.17.0.dev20240423.dist-info/RECORD,,