Avoid TimeDistributed layers to fix for keras 3.3.3 and Acknowledge r…

…andomness for test

keras_cv/src/models/object_detection/mask_rcnn/mask_head.py

@@ -49,49 +49,53 @@ def __init__(
         self.num_deconv_filters = num_deconv_filters
         self.layers = []
         for num_filters in stackwise_num_conv_filters:
-            conv = keras.layers.TimeDistributed(
-                keras.layers.Conv2D(
-                    filters=num_filters,
-                    kernel_size=3,
-                    padding="same",
-                )
-            )
-            batchnorm = keras.layers.TimeDistributed(
-                keras.layers.BatchNormalization()
+            conv = keras.layers.Conv2D(
+                filters=num_filters,
+                kernel_size=3,
+                padding="same",
             )
+            batchnorm = keras.layers.BatchNormalization()
             activation = keras.layers.Activation("relu")
             self.layers.extend([conv, batchnorm, activation])
 
-        self.deconv = keras.layers.TimeDistributed(
-            keras.layers.Conv2DTranspose(
-                num_deconv_filters,
-                kernel_size=2,
-                strides=2,
-                activation="relu",
-                padding="valid",
-            )
+        self.deconv = keras.layers.Conv2DTranspose(
+            num_deconv_filters,
+            kernel_size=2,
+            strides=2,
+            activation="relu",
+            padding="valid",
         )
         # we do not use a final sigmoid activation, since we use
         # from_logits=True during training
-        self.segmentation_mask_output = keras.layers.TimeDistributed(
-            keras.layers.Conv2D(
-                num_classes + 1,
-                kernel_size=1,
-                strides=1,
-                activation="linear",
-            )
+        self.segmentation_mask_output = keras.layers.Conv2D(
+            num_classes + 1,
+            kernel_size=1,
+            strides=1,
+            activation="linear",
         )
 
     def call(self, feature_map, training=False):
-        x = feature_map
+        # reshape batch and ROI axes into one axis to obtain a suitable
+        # shape for conv layers
+        num_rois = keras.ops.shape(feature_map)[1]
+        x = keras.ops.reshape(feature_map, (-1, *feature_map.shape[2:]))
         for layer in self.layers:
             x = layer(x, training=training)
         x = self.deconv(x)
-        mask = self.segmentation_mask_output(x)
-        return mask
+        segmentation_mask = self.segmentation_mask_output(x)
+        segmentation_mask = keras.ops.reshape(
+            segmentation_mask, (-1, num_rois, *segmentation_mask.shape[1:])
+        )
+        return segmentation_mask
 
     def build(self, input_shape):
-        intermediate_shape = input_shape
+        if input_shape[0] is None or input_shape[1] is None:
+            intermediate_shape = (None, *input_shape[2:])
+        else:
+            intermediate_shape = (
+                input_shape[0] * input_shape[1],
+                *input_shape[2:],
+            )
         for idx, num_filters in enumerate(self.stackwise_num_conv_filters):
             self.layers[idx * 3].build(intermediate_shape)
             intermediate_shape = tuple(intermediate_shape[:-1]) + (num_filters,)

keras_cv/src/models/object_detection/mask_rcnn/roi_sampler_test.py

@@ -76,8 +76,9 @@ def test_roi_sampler(self, mask_value):
         )
         # the sampled mask is only set to 1 if the ground truth
         # mask indicates object 2
+        sampled_index = ops.where(sampled_gt_classes[0, :, 0] == 10)[0][0]
         self.assertAllClose(
-            sampled_gt_masks[0, 0],
+            sampled_gt_masks[0, sampled_index],
             (mask_value == 2) * np.ones((14, 14)),
         )