How to add class to pretrained model?

[hunter.norman]

According to this whitepaper it is possible to add a new class without replacing the old classes while training.

My question is how do I modify the spec files to do this and what should my data look like? If I have a dataset that only has the new object and labels for it will that be ok? When I am modifying the spec files for re-training with TAO do I just add the new class alongside the others or should I remove the pre-existing classes from the spec files for re-training?

[Morganh]

You can refer to below.

When you train this new model to add a helmet class, you must provide a properly annotated dataset that has both people and helmets. If you only provide helmet data and not people data, the model performs well on helmets but poorly on people. Generally speaking, in transfer learning when adding new classes, you need representative data that covers the existing classes along with new classes.