How to use separate detection heads in ConcatDataset?

[jinga-lala]

Hi,
I want to use Concatenate Dataset option in MMDetection framework for Object detection? Particularly this,
https://mmdetection.readthedocs.io/en/latest/tutorials/customize_dataset.html#concatenate-dataset
I want to combine two supported datasets: COCO and VOC. However, both of them have different number of classes.
How can I ensure that the object detection heads are different for both models but we have same backbone?
Is there any other functionality in MMDetection that I can use to achieve this?

Thanks