An algorithm can predict how memorable or forgettable an image is almost as accurately as humans
You want to know how memorable your photos are? Don’t worry as a computer can tell you just that. An algorithm has been developed by MIT scientists that can almost accurately forecast how memorable or forgettable a picture is, just like human beings.
In a press release, Aditya Khosla, the lead author on a related paper, said, “Understanding memorability can help us make systems to capture 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 visual message”.
Actually, the new algorithm has been set to be developed into an app that delicately tweaks photos for making them more memorable. But, there are many other applications for the same.
For instance, the algorithm may be used for enhancement of the content of ads and posts on social media. The algorithm can be possibly helpful in the development of more efficient teaching resources and for the creation of a personal device to aid people in remembering stuff.
While developing it, the researchers in reality pitted the algorithm against human subjects and made the model foretell how memorable a team of people would find a new, never-before-seen photograph. The performance of the algorithm was 30% better as compared to existing algorithms. Its performance was within some percentage points of the average human performance.
Aude Oliva, principal research scientist said that while much progress has been propelled by deep-learning in recognizing object and understanding scene, human memory prediction has generally been considered a higher-level cognitive process that will never be tackled by computer scientists. Oliva added that such a thing could be done and they have done it.