Unable to freeze weights in object detection model

With reference to #2203 , I am trying to perform object detection with the ssd mobienet model and I am trying to freeze the weights.
Upon printing the layers with print(detection_models.trainable_variables) I get

Without freezing,
I get 232 trainable weights and 152 non trainable weights.
len(detection_model.trainable_weights), len(detection_model.non_trainable_weights)
I am trying to add the
freeze_variables: ".* ssd_mobile_net_v2fpn_keras_feature_extractor.* " and change the freeze_batchnorm parameter to true
tinthe train_config and then print the number of trainable weights.

I get the same values of 232 trainable weights and 152 non trainable weights.
len(detection_model.trainable_weights), len(detection_model.non_trainable_weights)

How can I check if the layer is actually frozen?