The more data we use to train a model (or as you said, the more patches we use), the better it’s performance will be.