Adding Differential Binarization model from PaddleOCR to Keras3

keras_hub/api/layers/__init__.py

@@ -41,6 +41,9 @@
 from keras_hub.src.models.densenet.densenet_image_converter import (
     DenseNetImageConverter,
 )
+from keras_hub.src.models.differential_binarization.differential_binarization_image_converter import (
+    DifferentialBinarizationImageConverter,
+)
 from keras_hub.src.models.efficientnet.efficientnet_image_converter import (
     EfficientNetImageConverter,
 )

keras_hub/api/models/__init__.py

@@ -104,6 +104,15 @@
 from keras_hub.src.models.densenet.densenet_image_classifier_preprocessor import (
     DenseNetImageClassifierPreprocessor,
 )
+from keras_hub.src.models.differential_binarization.differential_binarization_backbone import (
+    DifferentialBinarizationBackbone,
+)
+from keras_hub.src.models.differential_binarization.differential_binarization_ocr import (
+    DifferentialBinarizationOCR,
+)
+from keras_hub.src.models.differential_binarization.differential_binarization_preprocessor import (
+    DifferentialBinarizationPreprocessor,
+)
 from keras_hub.src.models.distil_bert.distil_bert_backbone import (
     DistilBertBackbone,
 )

keras_hub/src/models/differential_binarization/__init__.py

@@ -0,0 +1,9 @@
+from keras_hub.src.models.differential_binarization.differential_binarization_backbone import (
+    DifferentialBinarizationBackbone,
+)
+from keras_hub.src.models.differential_binarization.differential_binarization_presets import (
+    backbone_presets,
+)
+from keras_hub.src.utils.preset_utils import register_presets
+
+register_presets(backbone_presets, DifferentialBinarizationBackbone)

keras_hub/src/models/differential_binarization/differential_binarization_backbone.py

@@ -0,0 +1,219 @@
+import keras
+from keras import layers
+
+from keras_hub.src.api_export import keras_hub_export
+from keras_hub.src.models.backbone import Backbone
+
+
+@keras_hub_export("keras_hub.models.DifferentialBinarizationBackbone")
+class DifferentialBinarizationBackbone(Backbone):
+    """
+    A Keras model implementing the Differential Binarization
+    architecture for scene text detection, described in
+    [Real-time Scene Text Detection with Differentiable Binarization](
+    https://arxiv.org/abs/1911.08947).
+
+    This class contains the backbone architecture containing the feature
+    pyramid network and model heads.
+
+    Args:
+        image_encoder: A `keras_hub.models.ResNetBackbone` instance.
+        fpn_channels: int. The number of channels to output by the feature
+            pyramid network. Defaults to 256.
+        head_kernel_list: list of ints. The kernel sizes of probability map and
+            threshold map heads. Defaults to [3, 2, 2].
+        dtype: `None` or str or `keras.mixed_precision.DTypePolicy`. The dtype
+            to use for the model's computations and weights.
+    """
+
+    def __init__(
+        self,
+        image_encoder,
+        fpn_channels=256,
+        head_kernel_list=[3, 2, 2],
+        dtype=None,
+        **kwargs,
+    ):
+        # === Functional Model ===
+        inputs = image_encoder.input
+        x = image_encoder.pyramid_outputs
+        x = diffbin_fpn_model(x, out_channels=fpn_channels, dtype=dtype)
+
+        probability_maps = diffbin_head(
+            x,
+            in_channels=fpn_channels,
+            kernel_list=head_kernel_list,
+            name="head_prob",
+        )
+        threshold_maps = diffbin_head(
+            x,
+            in_channels=fpn_channels,
+            kernel_list=head_kernel_list,
+            name="head_thresh",
+        )
+
+        outputs = {
+            "probability_maps": probability_maps,
+            "threshold_maps": threshold_maps,
+        }
+
+        super().__init__(inputs=inputs, outputs=outputs, dtype=dtype, **kwargs)
+
+        # === Config ===
+        self.image_encoder = image_encoder
+        self.fpn_channels = fpn_channels
+        self.head_kernel_list = head_kernel_list
+
+    def get_config(self):
+        config = super().get_config()
+        config["fpn_channels"] = self.fpn_channels
+        config["head_kernel_list"] = self.head_kernel_list
+        config["image_encoder"] = keras.layers.serialize(self.image_encoder)
+        return config
+
+    @classmethod
+    def from_config(cls, config):
+        config["image_encoder"] = keras.layers.deserialize(
+            config["image_encoder"]
+        )
+        return cls(**config)
+
+
+def diffbin_fpn_model(inputs, out_channels, dtype=None):
+    # lateral layers composing the FPN's bottom-up pathway using
+    # pointwise convolutions of ResNet's pyramid outputs
+    lateral_p2 = layers.Conv2D(
+        out_channels,
+        kernel_size=1,
+        use_bias=False,
+        name="neck_lateral_p2",
+        dtype=dtype,
+    )(inputs["P2"])
+    lateral_p3 = layers.Conv2D(
+        out_channels,
+        kernel_size=1,
+        use_bias=False,
+        name="neck_lateral_p3",
+        dtype=dtype,
+    )(inputs["P3"])
+    lateral_p4 = layers.Conv2D(
+        out_channels,
+        kernel_size=1,
+        use_bias=False,
+        name="neck_lateral_p4",
+        dtype=dtype,
+    )(inputs["P4"])
+    lateral_p5 = layers.Conv2D(
+        out_channels,
+        kernel_size=1,
+        use_bias=False,
+        name="neck_lateral_p5",
+        dtype=dtype,
+    )(inputs["P5"])
+    # top-down fusion pathway consisting of upsampling layers with
+    # skip connections
+    topdown_p5 = lateral_p5
+    topdown_p4 = layers.Add(name="neck_topdown_p4")(
+        [
+            layers.UpSampling2D(dtype=dtype)(topdown_p5),
+            lateral_p4,
+        ]
+    )
+    topdown_p3 = layers.Add(name="neck_topdown_p3")(
+        [
+            layers.UpSampling2D(dtype=dtype)(topdown_p4),
+            lateral_p3,
+        ]
+    )
+    topdown_p2 = layers.Add(name="neck_topdown_p2")(
+        [
+            layers.UpSampling2D(dtype=dtype)(topdown_p3),
+            lateral_p2,
+        ]
+    )
+    # construct merged feature maps for each pyramid level
+    featuremap_p5 = layers.Conv2D(
+        out_channels // 4,
+        kernel_size=3,
+        padding="same",
+        use_bias=False,
+        name="neck_featuremap_p5",
+        dtype=dtype,
+    )(topdown_p5)
+    featuremap_p4 = layers.Conv2D(
+        out_channels // 4,
+        kernel_size=3,
+        padding="same",
+        use_bias=False,
+        name="neck_featuremap_p4",
+        dtype=dtype,
+    )(topdown_p4)
+    featuremap_p3 = layers.Conv2D(
+        out_channels // 4,
+        kernel_size=3,
+        padding="same",
+        use_bias=False,
+        name="neck_featuremap_p3",
+        dtype=dtype,
+    )(topdown_p3)
+    featuremap_p2 = layers.Conv2D(
+        out_channels // 4,
+        kernel_size=3,
+        padding="same",
+        use_bias=False,
+        name="neck_featuremap_p2",
+        dtype=dtype,
+    )(topdown_p2)
+    featuremap_p5 = layers.UpSampling2D((8, 8), dtype=dtype)(featuremap_p5)
+    featuremap_p4 = layers.UpSampling2D((4, 4), dtype=dtype)(featuremap_p4)
+    featuremap_p3 = layers.UpSampling2D((2, 2), dtype=dtype)(featuremap_p3)
+    featuremap = layers.Concatenate(axis=-1, dtype=dtype)(
+        [featuremap_p5, featuremap_p4, featuremap_p3, featuremap_p2]
+    )
+    return featuremap
+
+
+def diffbin_head(inputs, in_channels, kernel_list, name):
+    x = layers.Conv2D(
+        in_channels // 4,
+        kernel_size=kernel_list[0],
+        padding="same",
+        use_bias=False,
+        name=f"{name}_conv0_weights",
+    )(inputs)
+    x = layers.BatchNormalization(
+        beta_initializer=keras.initializers.Constant(1e-4),
+        gamma_initializer=keras.initializers.Constant(1.0),
+        name=f"{name}_conv0_bn",
+    )(x)
+    x = layers.ReLU(name=f"{name}_conv0_relu")(x)
+    x = layers.Conv2DTranspose(
+        in_channels // 4,
+        kernel_size=kernel_list[1],
+        strides=2,
+        padding="valid",
+        bias_initializer=keras.initializers.RandomUniform(
+            minval=-1.0 / (in_channels // 4 * 1.0) ** 0.5,
+            maxval=1.0 / (in_channels // 4 * 1.0) ** 0.5,
+        ),
+        name=f"{name}_conv1_weights",
+    )(x)
+    x = layers.BatchNormalization(
+        beta_initializer=keras.initializers.Constant(1e-4),
+        gamma_initializer=keras.initializers.Constant(1.0),
+        name=f"{name}_conv1_bn",
+    )(x)
+    x = layers.ReLU(name=f"{name}_conv1_relu")(x)
+    x = layers.Conv2DTranspose(
+        1,
+        kernel_size=kernel_list[2],
+        strides=2,
+        padding="valid",
+        activation="sigmoid",
+        bias_initializer=keras.initializers.RandomUniform(
+            minval=-1.0 / (in_channels // 4 * 1.0) ** 0.5,
+            maxval=1.0 / (in_channels // 4 * 1.0) ** 0.5,
+        ),
+        name=f"{name}_conv2_weights",
+    )(x)
+    return x

keras_hub/src/models/differential_binarization/differential_binarization_backbone_test.py

@@ -0,0 +1,44 @@
+from keras import ops
+
+from keras_hub.src.models.differential_binarization.differential_binarization_backbone import (
+    DifferentialBinarizationBackbone,
+)
+from keras_hub.src.models.differential_binarization.differential_binarization_preprocessor import (
+    DifferentialBinarizationPreprocessor,
+)
+from keras_hub.src.models.resnet.resnet_backbone import ResNetBackbone
+from keras_hub.src.tests.test_case import TestCase
+
+
+class DifferentialBinarizationTest(TestCase):
+    def setUp(self):
+        self.images = ops.ones((2, 32, 32, 3))
+        self.image_encoder = ResNetBackbone(
+            input_conv_filters=[4],
+            input_conv_kernel_sizes=[7],
+            stackwise_num_filters=[64, 4, 4, 4],
+            stackwise_num_blocks=[3, 4, 6, 3],
+            stackwise_num_strides=[1, 2, 2, 2],
+            block_type="bottleneck_block",
+            image_shape=(32, 32, 3),
+        )
+        self.preprocessor = DifferentialBinarizationPreprocessor()
+        self.init_kwargs = {
+            "image_encoder": self.image_encoder,
+            "fpn_channels": 16,
+            "head_kernel_list": [3, 2, 2],
+        }
+
+    def test_backbone_basics(self):
+        expected_output_shape = {
+            "probability_maps": (2, 32, 32, 1),
+            "threshold_maps": (2, 32, 32, 1),
+        }
+        self.run_backbone_test(
+            cls=DifferentialBinarizationBackbone,
+            init_kwargs=self.init_kwargs,
+            input_data=self.images,
+            expected_output_shape=expected_output_shape,
+            run_mixed_precision_check=False,
+            run_quantization_check=False,
+        )

keras_hub/src/models/differential_binarization/differential_binarization_image_converter.py

@@ -0,0 +1,10 @@
+from keras_hub.src.api_export import keras_hub_export
+from keras_hub.src.layers.preprocessing.image_converter import ImageConverter
+from keras_hub.src.models.differential_binarization.differential_binarization_backbone import (
+    DifferentialBinarizationBackbone,
+)
+
+
+@keras_hub_export("keras_hub.layers.DifferentialBinarizationImageConverter")
+class DifferentialBinarizationImageConverter(ImageConverter):
+    backbone_cls = DifferentialBinarizationBackbone