MIT Researchers Develop Algorithm to Predict Your Photo’s Memorability

MIT Researchers Develop Algorithm to Predict Your Photo’s Memorability

MIT scientists said they have developed a system capable of predicting the strength of human memory, thanks to artificial intelligence. They claimed their new algorithm tells accurately how well humans remember different pictures.

The researchers said they have created an algorithm, called the ‘MemNet’ algorithm, which is based on deep-learning techniques. It was exposed to a number of pictures, each with a score on how well people remembered them.

MemNet closely analyzed the pictures and determined a pattern to know what pictures people remembered and what areas of them contributed most to that.

After that, the researchers used some new pictures unaccompanied by a memorability score. They found that MemNet predicted a score very near to people’s score. Now, enthusiastic people can test the algorithm online. The system is capable of analyzing a picture that a user can upload and estimate its memorability. It shows a heat map of what items it has contributed most to it, the researchers said.

Aditya Khosla, lead author of the study of algorithm, said, “Understanding memorability can help us make systems to capture the most important information, or, conversely, to store information that humans will most likely forget. It's like having an instant focus group that tells you how likely it is that someone will remember a visual message”.

The developers of the MemNet system explained it in a paper that they presented at the International Conference on Computer Vision. The results that they presented were striking.


Share

Contact

Whether you want to ask us a question, would like to solve a problem, or just give us a suggestion, you’ll find many ways to contact us right here.

Email: editor@norcal.news

Phone: (916) 225-9835

Fax: (916) 225-9845

Newsletters

Subscribe and get the latest updates, news and more...