Global rename of classes to num_classes

README.md

@@ -88,7 +88,7 @@ test_dataset = test_dataset.batch(16).map(
 densenet = keras_cv.models.DenseNet121(
     include_rescaling=True,
     include_top=True,
-    classes=3
+    num_classes=3
 )
 densenet.compile(
     loss='categorical_crossentropy',

benchmarks/metrics/coco/mean_average_precision_bucket_performance.py

@@ -12,7 +12,9 @@
 from keras_cv.metrics import coco
 
 
-def produce_random_data(include_confidence=False, num_images=128, classes=20):
+def produce_random_data(
+    include_confidence=False, num_images=128, num_classes=20
+):
     """Generates a fake list of bounding boxes for use in this test.
 
     Returns:
@@ -23,7 +25,7 @@ def produce_random_data(include_confidence=False, num_images=128, classes=20):
     images = []
     for _ in range(num_images):
         num_boxes = math.floor(25 * random.uniform(0, 1))
-        classes_in_image = np.floor(np.random.rand(num_boxes, 1) * classes)
+        classes_in_image = np.floor(np.random.rand(num_boxes, 1) * num_classes)
         bboxes = np.random.rand(num_boxes, 4)
         boxes = np.concatenate([bboxes, classes_in_image], axis=-1)
         if include_confidence:

benchmarks/metrics/coco/mean_average_precision_performance.py

@@ -12,7 +12,9 @@
 from keras_cv.metrics import coco
 
 
-def produce_random_data(include_confidence=False, num_images=128, classes=20):
+def produce_random_data(
+    include_confidence=False, num_images=128, num_classes=20
+):
     """Generates a fake list of bounding boxes for use in this test.
 
     Returns:
@@ -23,7 +25,7 @@ def produce_random_data(include_confidence=False, num_images=128, classes=20):
     images = []
     for _ in range(num_images):
         num_boxes = math.floor(25 * random.uniform(0, 1))
-        classes_in_image = np.floor(np.random.rand(num_boxes, 1) * classes)
+        classes_in_image = np.floor(np.random.rand(num_boxes, 1) * num_classes)
         bboxes = np.random.rand(num_boxes, 4)
         boxes = np.concatenate([bboxes, classes_in_image], axis=-1)
         if include_confidence:

benchmarks/metrics/coco/recall_performance.py

@@ -12,7 +12,9 @@
 from keras_cv.metrics import coco
 
 
-def produce_random_data(include_confidence=False, num_images=128, classes=20):
+def produce_random_data(
+    include_confidence=False, num_images=128, num_classes=20
+):
     """Generates a fake list of bounding boxes for use in this test.
 
     Returns:
@@ -23,7 +25,7 @@ def produce_random_data(include_confidence=False, num_images=128, classes=20):
     images = []
     for _ in range(num_images):
         num_boxes = math.floor(25 * random.uniform(0, 1))
-        classes_in_image = np.floor(np.random.rand(num_boxes, 1) * classes)
+        classes_in_image = np.floor(np.random.rand(num_boxes, 1) * num_classes)
         bboxes = np.random.rand(num_boxes, 4)
         boxes = np.concatenate([bboxes, classes_in_image], axis=-1)
         if include_confidence:

examples/benchmarking/imagenet_v2.py

@@ -60,7 +60,7 @@
 model = model(
     include_rescaling=FLAGS.include_rescaling,
     include_top=True,
-    classes=1000,
+    num_classes=1000,
     input_shape=(224, 224, 3),
     weights=FLAGS.weights,
     **eval(FLAGS.model_kwargs),

examples/layers/preprocessing/classification/demo_utils.py

@@ -17,9 +17,9 @@
 import tensorflow_datasets as tfds
 
 
-def resize(image, label, img_size=(224, 224), classes=10):
+def resize(image, label, img_size=(224, 224), num_classes=10):
     image = tf.image.resize(image, img_size)
-    label = tf.one_hot(label, classes)
+    label = tf.one_hot(label, num_classes)
     return {"images": image, "labels": label}
 
 
@@ -32,11 +32,11 @@ def load_oxford_dataset(
     # Load dataset.
     data, ds_info = tfds.load(name, as_supervised=as_supervised, with_info=True)
     train_ds = data["train"]
-    classes = ds_info.features["label"].num_classes
+    num_classes = ds_info.features["label"].num_classes
 
     # Get tf dataset.
     train_ds = train_ds.map(
-        lambda x, y: resize(x, y, img_size=img_size, classes=classes)
+        lambda x, y: resize(x, y, img_size=img_size, num_classes=num_classes)
     ).batch(batch_size)
     return train_ds
 

examples/training/classification/imagenet/basic_training.py

@@ -127,7 +127,7 @@
 FLAGS = flags.FLAGS
 FLAGS(sys.argv)
 
-CLASSES = 1000
+NUM_CLASSES = 1000
 IMAGE_SIZE = (224, 224)
 REDUCE_ON_PLATEAU = "ReduceOnPlateau"
 COSINE_DECAY_WITH_WARMUP = "CosineDecayWithWarmup"
@@ -239,7 +239,7 @@ def augment(img, label):
     model = model(
         include_rescaling=True,
         include_top=True,
-        classes=CLASSES,
+        num_classes=NUM_CLASSES,
         input_shape=IMAGE_SIZE + (3,),
         **eval(FLAGS.model_kwargs),
     )

examples/training/contrastive/imagenet/simclr_training.py

@@ -70,7 +70,7 @@
 if FLAGS.model_name not in models.__dict__:
     raise ValueError(f"Invalid model name: {FLAGS.model_name}")
 
-CLASSES = 1000
+NUM_CLASSES = 1000
 IMAGE_SIZE = (224, 224)
 EPOCHS = 250
 
@@ -102,7 +102,7 @@
         augmenter=training.SimCLRAugmenter(
             value_range=(0, 255), target_size=IMAGE_SIZE
         ),
-        probe=layers.Dense(CLASSES, name="linear_probe"),
+        probe=layers.Dense(NUM_CLASSES, name="linear_probe"),
     )
 
     optimizer = optimizers.SGD(

examples/training/object_detection/pascal_voc/faster_rcnn.py

@@ -96,7 +96,7 @@
     #     include_top=False, weights="imagenet", include_rescaling=False
     # ).as_backbone()
     model = keras_cv.models.FasterRCNN(
-        classes=20, bounding_box_format="yxyx", backbone=backbone
+        num_classes=20, bounding_box_format="yxyx", backbone=backbone
     )
 
 

examples/training/object_detection/pascal_voc/retina_net.py

@@ -356,7 +356,7 @@ def pad_fn(images, bounding_boxes):
     # ).as_backbone()
     model = keras_cv.models.RetinaNet(
         # number of classes to be used in box classification
-        classes=20,
+        num_classes=20,
         # For more info on supported bounding box formats, visit
         # https://keras.io/api/keras_cv/bounding_box/
         bounding_box_format="xywh",

examples/training/semantic_segmentation/pascal_voc/basic_training.py

@@ -156,7 +156,7 @@ def proc_eval_fn(examples):
         weights="imagenet",
     )
     model = models.__dict__[FLAGS.model_name]
-    model = model(classes=21, backbone=backbone, **eval(FLAGS.model_kwargs))
+    model = model(num_classes=21, backbone=backbone, **eval(FLAGS.model_kwargs))
     optimizer = tf.keras.optimizers.SGD(
         learning_rate=lr_decay, momentum=0.9, clipnorm=10.0
     )

keras_cv/benchmarks/classification_training_benchmark_test.py

@@ -66,7 +66,7 @@ def _run_benchmark(self, app, strategy):
 
             model = app(
                 include_top=True,
-                classes=self.num_classes,
+                num_classes=self.num_classes,
                 input_shape=(56, 56, 1),
                 include_rescaling=True,
             )

keras_cv/callbacks/pycoco_callback_test.py

@@ -32,7 +32,7 @@ def cleanup_global_session(self):
 
     def test_model_fit_retinanet(self):
         model = keras_cv.models.RetinaNet(
-            classes=10,
+            num_classes=10,
             bounding_box_format="xywh",
             backbone=keras_cv.models.ResNet50V2Backbone().get_feature_extractor(),
         )
@@ -61,7 +61,7 @@ def test_model_fit_retinanet(self):
 
     def test_model_fit_rcnn(self):
         model = keras_cv.models.FasterRCNN(
-            classes=10,
+            num_classes=10,
             bounding_box_format="xywh",
             backbone=keras_cv.models.ResNet50V2Backbone().get_feature_extractor(),
         )

keras_cv/layers/object_detection/multi_class_non_max_suppression_test.py

@@ -20,8 +20,8 @@
 
 
 def decode_predictions_output_shapes():
-    classes = 10
-    predictions_shape = (8, 98208, 4 + classes)
+    num_classes = 10
+    predictions_shape = (8, 98208, 4 + num_classes)
 
     predictions = tf.random.stateless_uniform(
         shape=predictions_shape,

keras_cv/layers/object_detection_3d/center_net_label_encoder.py

@@ -331,7 +331,7 @@ class CenterNetLabelEncoder(tf.keras.layers.Layer):
       min_radius: minimum Gasussian radius in each dimension in meters.
       max_radius: maximum Gasussian radius in each dimension in meters.
       spatial_size: the x, y, z boundary of voxels
-      classes: number of object classes.
+      num_classes: number of object classes.
       top_k_heatmap: A sequence of integers, top k for each class. Can be None.
     """
 
@@ -341,7 +341,7 @@ def __init__(
         min_radius: Sequence[float],
         max_radius: Sequence[float],
         spatial_size: Sequence[float],
-        classes: int,
+        num_classes: int,
         top_k_heatmap: Sequence[int],
         **kwargs,
     ):
@@ -350,7 +350,7 @@ def __init__(
         self._min_radius = min_radius
         self._max_radius = max_radius
         self._spatial_size = spatial_size
-        self._classes = classes
+        self._num_classes = num_classes
         self._top_k_heatmap = top_k_heatmap
 
     def call(self, box_3d, box_classes, box_mask):
@@ -418,7 +418,7 @@ def call(self, box_3d, box_classes, box_mask):
         heatmap_dict = {}
         box_3d_dict = {}
         top_k_heatmap_feature_idx_dict = {}
-        for i in range(self._classes):
+        for i in range(self._num_classes):
             class_key = f"class_{i+1}"
             # Object class is 1-indexed (0 is background).
             dense_box_class_i = tf.cast(

keras_cv/layers/object_detection_3d/center_net_label_encoder_test.py

@@ -27,7 +27,7 @@ def test_voxelization_output_shape_no_z(self):
             min_radius=[0.8, 0.8, 0.0],
             max_radius=[8.0, 8.0, 0.0],
             spatial_size=[-20, 20, -20, 20, -20, 20],
-            classes=2,
+            num_classes=2,
             top_k_heatmap=[10, 20],
         )
         box_3d = tf.random.uniform(
@@ -53,7 +53,7 @@ def test_voxelization_output_shape_with_z(self):
             min_radius=[0.8, 0.8, 0.0],
             max_radius=[8.0, 8.0, 0.0],
             spatial_size=[-20, 20, -20, 20, -20, 20],
-            classes=2,
+            num_classes=2,
             top_k_heatmap=[10, 20],
         )
         box_3d = tf.random.uniform(
@@ -79,7 +79,7 @@ def test_voxelization_output_shape_missing_topk(self):
             min_radius=[0.8, 0.8, 0.0],
             max_radius=[8.0, 8.0, 0.0],
             spatial_size=[-20, 20, -20, 20, -20, 20],
-            classes=2,
+            num_classes=2,
             top_k_heatmap=[10, 0],
         )
         box_3d = tf.random.uniform(

keras_cv/layers/preprocessing/cut_mix_test.py

@@ -15,7 +15,7 @@
 
 from keras_cv.layers.preprocessing.cut_mix import CutMix
 
-classes = 10
+num_classes = 10
 
 
 class CutMixTest(tf.test.TestCase):
@@ -24,7 +24,7 @@ def test_return_shapes(self):
         # randomly sample labels
         ys = tf.random.categorical(tf.math.log([[0.5, 0.5]]), 2)
         ys = tf.squeeze(ys)
-        ys = tf.one_hot(ys, classes)
+        ys = tf.one_hot(ys, num_classes)
 
         layer = CutMix(seed=1)
         outputs = layer({"images": xs, "labels": ys})

keras_cv/layers/preprocessing/fourier_mix_test.py

@@ -15,7 +15,7 @@
 
 from keras_cv.layers.preprocessing.fourier_mix import FourierMix
 
-classes = 10
+num_classes = 10
 
 
 class FourierMixTest(tf.test.TestCase):
@@ -24,7 +24,7 @@ def test_return_shapes(self):
         # randomly sample labels
         ys = tf.random.categorical(tf.math.log([[0.5, 0.5]]), 2)
         ys = tf.squeeze(ys)
-        ys = tf.one_hot(ys, classes)
+        ys = tf.one_hot(ys, num_classes)
 
         layer = FourierMix()
         outputs = layer({"images": xs, "labels": ys})

keras_cv/layers/preprocessing/mix_up_test.py

@@ -15,7 +15,7 @@
 
 from keras_cv.layers.preprocessing.mix_up import MixUp
 
-classes = 10
+num_classes = 10
 
 
 class MixUpTest(tf.test.TestCase):
@@ -24,7 +24,7 @@ def test_return_shapes(self):
         # randomly sample labels
         ys_labels = tf.random.categorical(tf.math.log([[0.5, 0.5]]), 2)
         ys_labels = tf.squeeze(ys_labels)
-        ys_labels = tf.one_hot(ys_labels, classes)
+        ys_labels = tf.one_hot(ys_labels, num_classes)
 
         # randomly sample bounding boxes
         ys_bounding_boxes = {

keras_cv/layers/preprocessing/mosaic_test.py

@@ -15,7 +15,7 @@
 
 from keras_cv.layers.preprocessing.mosaic import Mosaic
 
-classes = 10
+num_classes = 10
 
 
 class MosaicTest(tf.test.TestCase):
@@ -24,7 +24,7 @@ def test_return_shapes(self):
         # randomly sample labels
         ys_labels = tf.random.categorical(tf.math.log([[0.5, 0.5]]), 2)
         ys_labels = tf.squeeze(ys_labels)
-        ys_labels = tf.one_hot(ys_labels, classes)
+        ys_labels = tf.one_hot(ys_labels, num_classes)
 
         # randomly sample bounding boxes
         ys_bounding_boxes = {

keras_cv/layers/preprocessing/random_crop_and_resize_test.py

@@ -127,12 +127,12 @@ def test_crop_area_factor_errors(self, crop_area_factor):
             )
 
     def test_augment_sparse_segmentation_mask(self):
-        classes = 8
+        num_classes = 8
 
         input_image_shape = (1, self.height, self.width, 3)
         mask_shape = (1, self.height, self.width, 1)
         image = tf.random.uniform(shape=input_image_shape, seed=self.seed)
-        mask = np.random.randint(2, size=mask_shape) * (classes - 1)
+        mask = np.random.randint(2, size=mask_shape) * (num_classes - 1)
 
         inputs = {"images": image, "segmentation_masks": mask}
 
@@ -160,16 +160,17 @@ def test_augment_sparse_segmentation_mask(self):
         self.assertAllInSet(output["segmentation_masks"], [0, 7])
 
     def test_augment_one_hot_segmentation_mask(self):
-        classes = 8
+        num_classes = 8
 
         input_image_shape = (1, self.height, self.width, 3)
         mask_shape = (1, self.height, self.width, 1)
         image = tf.random.uniform(shape=input_image_shape, seed=self.seed)
         mask = tf.one_hot(
             tf.squeeze(
-                np.random.randint(2, size=mask_shape) * (classes - 1), axis=-1
+                np.random.randint(2, size=mask_shape) * (num_classes - 1),
+                axis=-1,
             ),
-            classes,
+            num_classes,
         )
 
         inputs = {"images": image, "segmentation_masks": mask}