> you still need to consider every image Not necessarily. The classifier will not only tell you whether it sees a rat, but also its confidence. So you can manu

> you still need to consider every image Not necessarily. The classifier will not only tell you whether it sees a rat, but also its confidence. So you can manually classify just the low confidence images, retrain, and repeat until there's none left or you're happy with its detection rate.