A new algorithm called MemNet has been designed by MIT researchers and is capable of predicting the most memorable photos.
What is more, this actually has a much broader spectrum of applications than the obvious advantage of allowing users to take the best selfies.
Experts at the Computer Science and Artificial Intelligence Lab (CSAIL) of the Massachusetts Institute of Technology developed this program, which they later called MemNet.
Its selling point is that it can be almost as skilled as humans in determining how impactful a photograph is going to be, by assessing its emotional appeal and likelihood of being instantly recognizable.
According to developers, the set of instructions was created after analyzing a set of 60,000 images, stored in a data base nicknamed LaMem (Large-Scale Image Memorability).
Thanks to deep learning, the program was able to identify patterns in this incredibly large collection of data, and can now examine new visual information, accurately predicting its future impact.
For instance, until now researchers have discovered that faces and bare body parts are among the most memorable, while images depicting outdoor scenery tend to be more forgettable.
Experiments have shown that MemNet is approximately 30% more advanced than any other algorithms of this kind that had been created until now.
In addition, its performance is surpassed just by a small margin when measured against that of humans required to assess the memorability of a photograph.
Therefore, as its creators point out, the potential of this program can be tremendous, given what a significant role images play in today’s society.
On social media, users would be able to choose the photos that would generate the most significant response from their friends or followers.
Similarly, companies could employ this technology so as to create visually compelling logos and other elements required for building a strong organizational identity.
Aside from greatly simplifying decisions related to branding, MemNet could also boost a company’s marketing and advertising efforts.
For instance, being able to determine in advance which images are the most striking and instantly recognizable would result in creating much more effective and resonant print ads.
These would be much more easily remembered by potential buyers, therefore resulting in more significant revenues, while reducing the costs associated with market research.
As explained by Aditya Khosla, one of the graduate students who developed MemNet, another application would be in an academic setting, where teachers could be assisted in creating the most dazzling and attention-grabbing visual presentations, which students would be able to follow and recall without difficulty.
Even the medical sector could benefit from this tool, visually compelling images allowing people suffering from Alzheimer’s disease or other types of dementia to store and retain information more effectively.
Those wishing to test this newly invented algorithm can access a demonstration available online (at http://memorability.csail.mit.edu/demo.html). There, they will be able to upload an image, which will be assessed based on how easily remembered the program considers it to be.
The photo will also be carefully heat mapped so as to reveal the parts which are most likely to remain ingrained in the viewer’s memory: red portions will be the most striking ones, while blue ones will be those with the least emotional impact.
MIT researchers will provide more details regarding the algorithm at the International Conference on Computer Vision, taking place this week in Santiago, Chile.
They have also recently announced that they are currently working on developing an app which would incorporate MemNet technology, with the added benefit of actually providing users with tips regarding ways of enhancing memorability.
As the feature gains popularity, it might even come to be an indispensable part of photo editing software. Eventually, boosting the image’s appeal and impact might evolve into something as commonplace as cropping photos or airbrushing them.
Image Source: Pav Photography