Merge pull request #10338 from srihari-humbarwadi:readme

PiperOrigin-RevId: 413033276

official/vision/beta/serving/semantic_segmentation_test.py

@@ -31,8 +31,7 @@ from official.vision.beta.serving import semantic_segmentation
 class SemanticSegmentationExportTest(tf.test.TestCase, parameterized.TestCase):
 
   def _get_segmentation_module(self):
-    params = exp_factory.get_exp_config('seg_deeplabv3_pascal')
-    params.task.model.backbone.dilated_resnet.model_id = 50
+    params = exp_factory.get_exp_config('mnv2_deeplabv3_pascal')
     segmentation_module = semantic_segmentation.SegmentationModule(
         params, batch_size=1, input_image_size=[112, 112])
     return segmentation_module

official/vision/beta/tasks/maskrcnn.py

@@ -275,7 +275,9 @@ class MaskRCNNTask(base_task.Task):
             self._task_config.validation_data.input_path,
             self._task_config.validation_data.file_type,
             self._task_config.validation_data.num_examples,
-            self.task_config.model.include_mask, annotation_path)
+            self.task_config.model.include_mask, annotation_path,
+            regenerate_source_id=self._task_config.validation_data.decoder
+            .simple_decoder.regenerate_source_id)
       self.coco_metric = coco_evaluator.COCOEvaluator(
           annotation_file=annotation_path,
           include_mask=self._task_config.model.include_mask,

official/vision/beta/tasks/retinanet.py

@@ -13,14 +13,14 @@
 # limitations under the License.
 
 """RetinaNet task definition."""
-from typing import Any, Optional, List, Tuple, Mapping
+from typing import Any, List, Mapping, Optional, Tuple
 
 from absl import logging
 import tensorflow as tf
+
 from official.common import dataset_fn
 from official.core import base_task
 from official.core import task_factory
-from official.vision import keras_cv
 from official.vision.beta.configs import retinanet as exp_cfg
 from official.vision.beta.dataloaders import input_reader_factory
 from official.vision.beta.dataloaders import retinanet_input
@@ -28,6 +28,8 @@ from official.vision.beta.dataloaders import tf_example_decoder
 from official.vision.beta.dataloaders import tfds_factory
 from official.vision.beta.dataloaders import tf_example_label_map_decoder
 from official.vision.beta.evaluation import coco_evaluator
+from official.vision.beta.losses import focal_loss
+from official.vision.beta.losses import loss_utils
 from official.vision.beta.modeling import factory
 
 
@@ -155,9 +157,9 @@ class RetinaNetTask(base_task.Task):
       if head.name not in outputs['attribute_outputs']:
         raise ValueError(f'Attribute {head.name} not found in model outputs.')
 
-      y_true_att = keras_cv.losses.multi_level_flatten(
+      y_true_att = loss_utils.multi_level_flatten(
           labels['attribute_targets'][head.name], last_dim=head.size)
-      y_pred_att = keras_cv.losses.multi_level_flatten(
+      y_pred_att = loss_utils.multi_level_flatten(
           outputs['attribute_outputs'][head.name], last_dim=head.size)
       if head.type == 'regression':
         att_loss_fn = tf.keras.losses.Huber(
@@ -180,7 +182,7 @@ class RetinaNetTask(base_task.Task):
     params = self.task_config
     attribute_heads = self.task_config.model.head.attribute_heads
 
-    cls_loss_fn = keras_cv.losses.FocalLoss(
+    cls_loss_fn = focal_loss.FocalLoss(
         alpha=params.losses.focal_loss_alpha,
         gamma=params.losses.focal_loss_gamma,
         reduction=tf.keras.losses.Reduction.SUM)
@@ -194,14 +196,14 @@ class RetinaNetTask(base_task.Task):
     num_positives = tf.reduce_sum(box_sample_weight) + 1.0
     cls_sample_weight = cls_sample_weight / num_positives
     box_sample_weight = box_sample_weight / num_positives
-    y_true_cls = keras_cv.losses.multi_level_flatten(
+    y_true_cls = loss_utils.multi_level_flatten(
         labels['cls_targets'], last_dim=None)
     y_true_cls = tf.one_hot(y_true_cls, params.model.num_classes)
-    y_pred_cls = keras_cv.losses.multi_level_flatten(
+    y_pred_cls = loss_utils.multi_level_flatten(
         outputs['cls_outputs'], last_dim=params.model.num_classes)
-    y_true_box = keras_cv.losses.multi_level_flatten(
+    y_true_box = loss_utils.multi_level_flatten(
         labels['box_targets'], last_dim=4)
-    y_pred_box = keras_cv.losses.multi_level_flatten(
+    y_pred_box = loss_utils.multi_level_flatten(
         outputs['box_outputs'], last_dim=4)
 
     cls_loss = cls_loss_fn(

official/vision/detection/dataloader/anchor.py

@@ -21,13 +21,13 @@ from __future__ import print_function
 import collections
 
 import tensorflow as tf
-from official.vision import keras_cv
+from official.vision.beta.ops import iou_similarity
 from official.vision.detection.utils import box_utils
-from official.vision.detection.utils.object_detection import argmax_matcher
-from official.vision.detection.utils.object_detection import balanced_positive_negative_sampler
-from official.vision.detection.utils.object_detection import box_list
-from official.vision.detection.utils.object_detection import faster_rcnn_box_coder
-from official.vision.detection.utils.object_detection import target_assigner
+from official.vision.utils.object_detection import argmax_matcher
+from official.vision.utils.object_detection import balanced_positive_negative_sampler
+from official.vision.utils.object_detection import box_list
+from official.vision.utils.object_detection import faster_rcnn_box_coder
+from official.vision.utils.object_detection import target_assigner
 
 
 class Anchor(object):
@@ -135,7 +135,7 @@ class AnchorLabeler(object):
         upper-bound threshold to assign negative labels for anchors. An anchor
         with a score below the threshold is labeled negative.
     """
-    similarity_calc = keras_cv.ops.IouSimilarity()
+    similarity_calc = iou_similarity.IouSimilarity()
     matcher = argmax_matcher.ArgMaxMatcher(
         match_threshold,
         unmatched_threshold=unmatched_threshold,
@@ -341,7 +341,7 @@ class OlnAnchorLabeler(RpnAnchorLabeler):
         unmatched_threshold=unmatched_threshold,
         rpn_batch_size_per_im=rpn_batch_size_per_im,
         rpn_fg_fraction=rpn_fg_fraction)
-    similarity_calc = keras_cv.ops.IouSimilarity()
+    similarity_calc = iou_similarity.IouSimilarity()
     matcher = argmax_matcher.ArgMaxMatcher(
         match_threshold,
         unmatched_threshold=unmatched_threshold,

official/vision/detection/dataloader/input_reader.py

@@ -16,7 +16,6 @@
 
 from __future__ import absolute_import
 from __future__ import division
-# from __future__ import google_type_annotations
 from __future__ import print_function
 
 import tensorflow as tf

official/vision/detection/executor/detection_executor.py

@@ -16,14 +16,13 @@
 
 from __future__ import absolute_import
 from __future__ import division
-# from __future__ import google_type_annotations
 from __future__ import print_function
 
 from absl import logging
 
 import tensorflow as tf
 from official.vision.detection.executor import distributed_executor as executor
-from official.vision.detection.utils.object_detection import visualization_utils
+from official.vision.utils.object_detection import visualization_utils
 
 
 class DetectionDistributedExecutor(executor.DistributedExecutor):

official/vision/detection/executor/distributed_executor.py

@@ -16,7 +16,6 @@
 
 from __future__ import absolute_import
 from __future__ import division
-# from __future__ import google_type_annotations
 from __future__ import print_function
 
 import os

official/vision/detection/modeling/architecture/nn_blocks.py

@@ -23,7 +23,6 @@ import tensorflow as tf
 from official.modeling import tf_utils
 
 
-@tf.keras.utils.register_keras_serializable(package='Vision')
 class ResidualBlock(tf.keras.layers.Layer):
   """A residual block."""
 
@@ -163,7 +162,6 @@ class ResidualBlock(tf.keras.layers.Layer):
     return self._activation_fn(x + shortcut)
 
 
-@tf.keras.utils.register_keras_serializable(package='Vision')
 class BottleneckBlock(tf.keras.layers.Layer):
   """A standard bottleneck block."""
 

official/vision/detection/modeling/architecture/spinenet.py

@@ -113,7 +113,6 @@ def build_block_specs(block_specs=None):
   return [BlockSpec(*b) for b in block_specs]
 
 
-@tf.keras.utils.register_keras_serializable(package='Vision')
 class SpineNet(tf.keras.Model):
   """Class to build SpineNet models."""
 

official/vision/detection/ops/target_ops.py

@@ -22,7 +22,7 @@ import tensorflow as tf
 
 from official.vision.detection.ops import spatial_transform_ops
 from official.vision.detection.utils import box_utils
-from official.vision.detection.utils.object_detection import balanced_positive_negative_sampler
+from official.vision.utils.object_detection import balanced_positive_negative_sampler
 
 
 def box_matching(boxes, gt_boxes, gt_classes):

official/vision/detection/utils/box_utils.py

@@ -16,7 +16,6 @@
 
 from __future__ import absolute_import
 from __future__ import division
-# from __future__ import google_type_annotations
 from __future__ import print_function
 
 import numpy as np

official/vision/detection/utils/input_utils.py

@@ -19,7 +19,7 @@ import math
 import tensorflow as tf
 
 from official.vision.detection.utils import box_utils
-from official.vision.detection.utils.object_detection import preprocessor
+from official.vision.utils.object_detection import preprocessor
 
 
 def pad_to_fixed_size(input_tensor, size, constant_values=0):

official/vision/image_classification/augment.py

@@ -20,7 +20,6 @@ RandAugment Reference: https://arxiv.org/abs/1909.13719
 
 from __future__ import absolute_import
 from __future__ import division
-# from __future__ import google_type_annotations
 from __future__ import print_function
 
 import math

official/vision/image_classification/augment_test.py

@@ -16,7 +16,6 @@
 
 from __future__ import absolute_import
 from __future__ import division
-# from __future__ import google_type_annotations
 from __future__ import print_function
 
 from absl.testing import parameterized

official/vision/image_classification/callbacks.py

@@ -16,7 +16,6 @@
 """Common modules for callbacks."""
 from __future__ import absolute_import
 from __future__ import division
-# from __future__ import google_type_annotations
 from __future__ import print_function
 
 import os

official/vision/image_classification/configs/base_configs.py

@@ -14,19 +14,49 @@
 
 # Lint as: python3
 """Definitions for high level configuration groups.."""
-from typing import Any, List, Mapping, Optional
 
 import dataclasses
-
+from typing import Any, List, Mapping, Optional
 from official.core import config_definitions
 from official.modeling import hyperparams
-from official.modeling.hyperparams import config_definitions as legacy_cfg
 
-CallbacksConfig = legacy_cfg.CallbacksConfig
-TensorboardConfig = legacy_cfg.TensorboardConfig
 RuntimeConfig = config_definitions.RuntimeConfig
 
 
+@dataclasses.dataclass
+class TensorBoardConfig(hyperparams.Config):
+  """Configuration for TensorBoard.
+
+  Attributes:
+    track_lr: Whether or not to track the learning rate in TensorBoard. Defaults
+      to True.
+    write_model_weights: Whether or not to write the model weights as images in
+      TensorBoard. Defaults to False.
+  """
+  track_lr: bool = True
+  write_model_weights: bool = False
+
+
+@dataclasses.dataclass
+class CallbacksConfig(hyperparams.Config):
+  """Configuration for Callbacks.
+
+  Attributes:
+    enable_checkpoint_and_export: Whether or not to enable checkpoints as a
+      Callback. Defaults to True.
+    enable_backup_and_restore: Whether or not to add BackupAndRestore
+      callback. Defaults to True.
+    enable_tensorboard: Whether or not to enable TensorBoard as a Callback.
+      Defaults to True.
+    enable_time_history: Whether or not to enable TimeHistory Callbacks.
+      Defaults to True.
+  """
+  enable_checkpoint_and_export: bool = True
+  enable_backup_and_restore: bool = False
+  enable_tensorboard: bool = True
+  enable_time_history: bool = True
+
+
 @dataclasses.dataclass
 class ExportConfig(hyperparams.Config):
   """Configuration for exports.
@@ -74,7 +104,7 @@ class TrainConfig(hyperparams.Config):
       inferred based on the number of images and batch size. Defaults to None.
     callbacks: An instance of CallbacksConfig.
     metrics: An instance of MetricsConfig.
-    tensorboard: An instance of TensorboardConfig.
+    tensorboard: An instance of TensorBoardConfig.
     set_epoch_loop: Whether or not to set `steps_per_execution` to
       equal the number of training steps in `model.compile`. This reduces the
       number of callbacks run per epoch which significantly improves end-to-end
@@ -85,7 +115,7 @@ class TrainConfig(hyperparams.Config):
   steps: int = None
   callbacks: CallbacksConfig = CallbacksConfig()
   metrics: MetricsConfig = None
-  tensorboard: TensorboardConfig = TensorboardConfig()
+  tensorboard: TensorBoardConfig = TensorBoardConfig()
   time_history: TimeHistoryConfig = TimeHistoryConfig()
   set_epoch_loop: bool = False
 

official/vision/image_classification/configs/configs.py

@@ -52,7 +52,7 @@ class EfficientNetImageNetConfig(base_configs.ExperimentConfig):
           enable_checkpoint_and_export=True, enable_tensorboard=True),
       metrics=['accuracy', 'top_5'],
       time_history=base_configs.TimeHistoryConfig(log_steps=100),
-      tensorboard=base_configs.TensorboardConfig(
+      tensorboard=base_configs.TensorBoardConfig(
           track_lr=True, write_model_weights=False),
       set_epoch_loop=False)
   evaluation: base_configs.EvalConfig = base_configs.EvalConfig(
@@ -84,7 +84,7 @@ class ResNetImagenetConfig(base_configs.ExperimentConfig):
           enable_checkpoint_and_export=True, enable_tensorboard=True),
       metrics=['accuracy', 'top_5'],
       time_history=base_configs.TimeHistoryConfig(log_steps=100),
-      tensorboard=base_configs.TensorboardConfig(
+      tensorboard=base_configs.TensorBoardConfig(
           track_lr=True, write_model_weights=False),
       set_epoch_loop=False)
   evaluation: base_configs.EvalConfig = base_configs.EvalConfig(

official/vision/image_classification/dataset_factory.py

@@ -16,7 +16,6 @@
 """Dataset utilities for vision tasks using TFDS and tf.data.Dataset."""
 from __future__ import absolute_import
 from __future__ import division
-# from __future__ import google_type_annotations
 from __future__ import print_function
 
 import os

official/vision/image_classification/efficientnet/common_modules.py

@@ -15,7 +15,6 @@
 """Common modeling utilities."""
 from __future__ import absolute_import
 from __future__ import division
-# from __future__ import google_type_annotations
 from __future__ import print_function
 
 import numpy as np

official/vision/image_classification/efficientnet/tfhub_export.py

@@ -16,7 +16,6 @@
 
 from __future__ import absolute_import
 from __future__ import division
-# from __future__ import google_type_annotations
 from __future__ import print_function
 
 import os