MIT’s new deep learning tool aims to predicts the memorability of your photos

Dec16

by Sindy Cator on December 16, 2015 at 10:14 pm

MIT computer scientists have developed a new algorithm that helps determine which parts of your photos are most memorable. The program, called MemNet, uses artificial intelligence to build a heat map which determines the most visually interesting pieces of your photo. Deep learning algorithms like MemNet use human training in order to “train” the algorithm how to complete complex processes. Over time, the algorithm gets smarter and can perform these tasks as well, or better, than a human. To train MemNet, researchers gathered tens of thousands of pictures from several datasets. Each picture was then assigned a memorability score based…