Add point to convert_keras_models.py in ResNet error message.

PiperOrigin-RevId: 324052434

research/object_detection/models/center_net_resnet_feature_extractor.py

@@ -129,8 +129,10 @@ class CenterNetResnetFeatureExtractor(CenterNetFeatureExtractor):
     else:
       supported_types = ['classification']
       raise ValueError(
-          ('Sub model {} is not defined for ResNet.'.format(sub_model_type) +
-           'Supported types are {}.'.format(supported_types)))
+          ('Sub model {} is not defined for ResNet.'.format(sub_model_type)
+           + 'Supported types are {}.'.format(supported_types)
+           + 'Use the script convert_keras_models.py to create your own '
+           + 'classification checkpoints.'))
 
 
 def resnet_v2_101(channel_means, channel_stds, bgr_ordering):

research/object_detection/models/center_net_resnet_v1_fpn_feature_extractor.py

@@ -166,7 +166,9 @@ class CenterNetResnetV1FpnFeatureExtractor(CenterNetFeatureExtractor):
       supported_types = ['classification']
       raise ValueError(
           ('Sub model {} is not defined for ResNet FPN.'.format(sub_model_type)
-           + 'Supported types are {}.'.format(supported_types)))
+           + 'Supported types are {}.'.format(supported_types))
+          + 'Use the script convert_keras_models.py to create your own '
+          + 'classification checkpoints.')
 
 
 def resnet_v1_101_fpn(channel_means, channel_stds, bgr_ordering):