Suppose you have two images and you want to know if they're the "same" image. One image might be cropped, scaled, saturated, etc., but a human could still quickly tell they're the same image. How could you get a machine to do it?

Here are a bunch of algorithms and their performance on 100 test samples. It measures the algorithm's success rate on both positive samples and negative samples.

The algorithms were tuned to prevent false positives (i.e. errors when dealing with negative samples). This may or may not be desired, depending on use case.