CenterNetResnetFeatureExtractor's 'supported_sub_model_types' property is set to 'classification'

In snippet below:
https://github.com/tensorflow/models/blob/7beddae1ff7207e7738693cdcdec389d16be83d3/research/object_detection/models/center_net_resnet_feature_extractor.py#L129-L130 the property is set as classification.

Fine-Tuning from CenterNet Resnet50 V2 512x512 I get the following error:
ValueError: Checkpoint type "detection" not supported for CenterNetResnetFeatureExtractor. Supported types are ['classification', 'fine_tune']

Config pipeline has been updated as: fine_tune_checkpoint_type: "detection"

Question is: Isn't CenterNet Resnet50 V2 512x512 supposed to be set as a detector ? Is there a difference in the api version between the pretrained model in the zoo and the current OD api ?