Don’t you feel jealous that some people happen to have faces that are unforgettable? Most of the time, these faces are the ones that you can stare at all day long without getting tired, i.e. being very beautiful or handsome. Still, if you were wondering why that is so, mathematics might break it down into something that is a whole lot more simple – and a team of researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) has done just that with the construction of MemNet. MemNet happens to be a deep learning based algorithm which can predict the “memorability” of your photographs, in a manner that is almost on par with that of the human brain.
Basically, the premise works this way – folks more or less remember and forget the same set of images across the board, where despite our different backgrounds and experiences, human brains still remain wired in similar manner – as we tend to remember a similar set of images.
MemNet makes use of Convolutional Neural Networks in order to perform predictions which are more or less as accurate as the memorability scores that hail from a wide range of humans. An MIT report even pointed out that it performed 30% better compared to current software, and was a few percentage points short of of the human average scores. You can be darn sure that folks would find different ways to monetize this for sure in the future!