It might. My understanding is, it is common to take trained models from one problem and use parts of them on other problems in a similar domain to speed up training.