Just re-train your own model on your own collected data. No problem there.