Combining tensorflow inception models/graphs

Question

I have done my own training of tensorflow inception a while ago, recently I wanted to add new classes to the original model. However, when I retrain tensorflow it generates a new graph and labels files completely separate from the older one.

Is there a way to combine models? Or do I have to retrain again with the old and new classes?

Answer 1

You have to retrain. You may be able to use transfer learning to do it much faster than you did before.