It’s a match, with fatchd!

fatchd is the new matching app using artificial intelligence.

What if you could use an app to find out before a date whether this two should meetor shouldn’t? All it takes is two facial images. After four seconds, it says: “You are 29% fatchd” or “You are 84% fatchd” (face matched).

This is how it works. No matter whether you are dealing with a celebrity, the Chancellor, your best friend, your colleagues or the next Tinder-Date matches, fatchd gives in an entertaining way a result within seconds. The matching value rotates in a circle from 0 to a maximum of 100% and is commented humorously, e.g. with “suffering preprogrammed”

My colleague Martin and I had a hypothesis: Is it possible to make a prediction from just two facial images that tells us whether two people match? Different approaches such as face reading methods or biometric algorithms did not lead to a feasible result.

The breakthrough came rather unexpected, when we confronted a scientific team from the Institute of Medical Engineering Systems at the Hamburg University of Technology with our work. They wanted to tackle the task from exactly the other side – what would an artificial intelligence say to all our data we have collected? After months of detailed work and hundreds of training and testing runs, light appeared at the end of the tunnel.

A machine learning model has been developed, which allows the “matching” of two people. If the problem is reduced to the question good or bad match (binary classification), the model achieves an accuracy of about 70%, which is much better than a normal human being can do. Thus, the scientific team has found substantial evidence that machine learning can be used for learning matching scores from the face images of two people. For this reason we have developed fatchd and made it available to everyone as a free game.

Let´s get fatchd!

More info: fatchd.com

Face Matched by Artificial Intelligence

Face Matched by Artificial Intelligence

Games

View all games

Anyone can write on Bored Panda. Start writing!

Follow Bored Panda on Google News!

Follow us on Flipboard.com/@boredpanda!