merge

research/object_detection/models/faster_rcnn_resnet_v1_fpn_keras_feature_extractor.py

@@ -20,6 +20,7 @@ import tensorflow.compat.v1 as tf
 from object_detection.meta_architectures import faster_rcnn_meta_arch
 from object_detection.models import feature_map_generators
 from object_detection.models.keras_models import resnet_v1
+from object_detection.utils import ops
 
 
 _RESNET_MODEL_OUTPUT_LAYERS = {
@@ -31,6 +32,49 @@ _RESNET_MODEL_OUTPUT_LAYERS = {
                       'conv4_block36_out', 'conv5_block3_out'],
 }
 
+class ResnetFPN(tf.keras.layers.Layer):
+    def __init__(self,
+                 backbone_classifier,
+                 fpn_features_generator,
+                 coarse_feature_layers,
+                 fpn_min_level,
+                 resnet_block_names,
+                 base_fpn_max_level):
+      super(ResnetFPN, self).__init__()
+      self.classification_backbone = backbone_classifier
+      self.fpn_features_generator = fpn_features_generator
+      self.coarse_feature_layers = coarse_feature_layers
+      self._fpn_min_level = fpn_min_level
+      self._resnet_block_names = resnet_block_names
+      self._base_fpn_max_level = base_fpn_max_level
+
+    def call(self, inputs):
+      inputs = ops.pad_to_multiple(inputs, 32)
+      backbone_outputs = self.classification_backbone(inputs)
+
+      feature_block_list = []
+      for level in range(self._fpn_min_level, self._base_fpn_max_level + 1):
+        feature_block_list.append('block{}'.format(level - 1))
+      feature_block_map = dict(
+          list(zip(self._resnet_block_names, backbone_outputs)))
+      fpn_input_image_features = [
+          (feature_block, feature_block_map[feature_block])
+          for feature_block in feature_block_list]
+      fpn_features = self.fpn_features_generator(fpn_input_image_features)
+
+      feature_maps = []
+      for level in range(self._fpn_min_level, self._base_fpn_max_level + 1):
+        feature_maps.append(fpn_features['top_down_block{}'.format(level-1)])
+      last_feature_map = fpn_features['top_down_block{}'.format(
+          self._base_fpn_max_level - 1)]
+
+      for coarse_feature_layers in self.coarse_feature_layers:
+        for layer in coarse_feature_layers:
+          last_feature_map = layer(last_feature_map)
+        feature_maps.append(last_feature_map)
+
+      return feature_maps
+
 
 class FasterRCNNResnetV1FpnKerasFeatureExtractor(
     faster_rcnn_meta_arch.FasterRCNNKerasFeatureExtractor):
@@ -155,9 +199,7 @@ class FasterRCNNResnetV1FpnKerasFeatureExtractor(
         self.classification_backbone = tf.keras.Model(
             inputs=full_resnet_v1_model.inputs,
             outputs=outputs)
-        backbone_outputs = self.classification_backbone(full_resnet_v1_model.inputs)
 
-        # construct FPN feature generator
         self._base_fpn_max_level = min(self._fpn_max_level, 5)
         self._num_levels = self._base_fpn_max_level + 1 - self._fpn_min_level
         self._fpn_features_generator = (
@@ -168,17 +210,7 @@ class FasterRCNNResnetV1FpnKerasFeatureExtractor(
                 conv_hyperparams=self._conv_hyperparams,
                 freeze_batchnorm=self._freeze_batchnorm,
                 name='FeatureMaps'))
-
-        feature_block_list = []
-        for level in range(self._fpn_min_level, self._base_fpn_max_level + 1):
-          feature_block_list.append('block{}'.format(level - 1))
-        feature_block_map = dict(
-            list(zip(self._resnet_block_names, backbone_outputs)))
-        fpn_input_image_features = [
-            (feature_block, feature_block_map[feature_block])
-            for feature_block in feature_block_list]
-        fpn_features = self._fpn_features_generator(fpn_input_image_features)
-
+        
         # Construct coarse feature layers
         for i in range(self._base_fpn_max_level, self._fpn_max_level):
           layers = []
@@ -199,20 +231,13 @@ class FasterRCNNResnetV1FpnKerasFeatureExtractor(
               self._conv_hyperparams.build_activation_layer(
                   name=layer_name))
           self._coarse_feature_layers.append(layers)
-
-        feature_maps = []
-        for level in range(self._fpn_min_level, self._base_fpn_max_level + 1):
-          feature_maps.append(fpn_features['top_down_block{}'.format(level-1)])
-        last_feature_map = fpn_features['top_down_block{}'.format(
-            self._base_fpn_max_level - 1)]
-
-        for coarse_feature_layers in self._coarse_feature_layers:
-          for layer in coarse_feature_layers:
-            last_feature_map = layer(last_feature_map)
-          feature_maps.append(last_feature_map)
-
-        feature_extractor_model = tf.keras.models.Model(
-            inputs=full_resnet_v1_model.inputs, outputs=feature_maps)
+        
+        feature_extractor_model = ResnetFPN(self.classification_backbone,
+                           self._fpn_features_generator,
+                           self._coarse_feature_layers,
+                           self._fpn_min_level,
+                           self._resnet_block_names,
+                           self._base_fpn_max_level)
         return feature_extractor_model
 
   def get_box_classifier_feature_extractor_model(self, name=None):