“You’re lying—I can see it on your face!” How many times have you scoffed at that expression? Thanks to new software developed by researchers from MIT’s Computer Science and Artificial Intelligence Laboratory, the next time you hear that, it might be the truth!
The software, called Eulerian Video Magnification (EVM), identifies temporal changes in color and movement—imperceptible to the human eye—in the individual pixels of videos. EVM then amplifies these changes so they can be seen more easily.
Let’s take a real-life example. Every time your heart beats, blood is pumped through your body. This process causes very subtle changes in your skin’s color as your blood vessels expand and contract. While you may not be able to see the difference in your skin’s color, EVM recognizes these minute changes and exaggerates them. In a manipulated video, your face would flash bright pink with every heartbeat, essentially making your pulse visible.
EVM was originally designed to monitor the vitals of neo-natal infants without physically touching them, but new uses have been suggested for law enforcement, construction management, and even gambling. Just imagine how a game of poker would change if you could easily call your opponent’s bluff.
CSAIL students Michael Rubinstein and Neal Wadhwa and MIT alumni Eugene Shih SM ’01, PhD ’10 and Hao-Yu Wu ’12, MNG ’12, along with Professor Frédo Durand, Professor William T. Freeman PhD ’92, and Professor John Guttag received an honorable mention from the National Science Foundation at the 10th annual International Science & Technology Visualization Challenge for their work on the video Revealing Invisible Changes In The World (below).
Interested in putting EVM to the test yourself? Taiwan-based Quanta Research Cambridge, a laptop computer manufacture and funder of this project, has created a platform that allows you to upload your videos and run them through the software. If you have Matlab, download EVM and try it on your own.