Learning an Imagenet Like Way Is Not Enough

Image net is just set up with these very strong correlations in a way that maybe human experience isn't, because we see a lot of cartoon lions as well as real lions. So i think if want to generalize well to the imagenet attest tat, you might actually not want to solve the problem in the same way as humans,. Because there is real signal in those textures that humans are ignoring slightly to their detriment. It doesn't incourage learning an athumman like way. Versis not. If all you care about is you're writing a phone ap, and you want to recognize something in them, like flowers, it's probably good to not be trying to do things