Imagine if all of these are carefully annotated and digitally imaged and assembled into a museum object corpus for searching but also for training some large multimodal models.