Maybe you are not a photography hobbyist or a professional directly so that you have been able to use programs of photo editing of great quality such as PhotoShop and the like. Even so, surely you have realized how difficult it can be to be able to edit a shot, taken for example from your smartphone, and capture that idea that surrounds you by using a series of editable filters, something that can be very simple or tremendously complex if we want something special.
With this in mind, perhaps much more interesting is the new milestone that researchers and engineers from Google with their new artificial intelligence systems since they have achieved that, in a completely autonomous way, a computer is capable of imitating the capabilities of a professional photographer in record time and offering truly impressive results.
Google presents its latest advances in artificial intelligence
Personally, I have to confess that something as simple as that Google does not want to commercialize this idea has caught my attention, at least for the moment. As has been officially commented, apparently we are only facing a experiment where it was sought to see what this new system of artificial intelligence capable of editing landscape photos at a level that has even deceived photographers who have dedicated a large part of their lives to this work.
Apparently, as you have commented Hui fang, a software engineer who works within the Google Machine Perception team, the true objective of this work was to demonstrate that artificial intelligence systems are not only used for tasks of 0 or 1, that is, to answer Yes or No to different issues, but may also be able to be trained to differentiate aesthetic content and perform much more subjective activities related to fields where their presence was not very common until now, such as art or photography.
In order to train the system, techniques of machine learning. In case you do not know well how these types of techniques work, tell you, in a very basic way, that thousands of photographs taken from Street View have been used to make the artificial intelligence system capable of detect panoramic landscapes to later be able to be edited following a photographer's workflow. The final objective that was pursued in this work was that the final result was pleasing to the human eye.
This project even manages to fool professional photographers
Once the functionality to be pursued by the system was defined, the engineers got to work and the result has been a software capable of choosing several photographs, following a series of established patterns to later crop them, adjust the lighting and saturation and present the result. The most important part of all this is found in that this artificial intelligence system can adjust these parameters by zones so it is not just a matter of applying a specific filter.
Once interesting results began to be obtained, you can see several of them distributed by this same entry or in the gallery that is located just below these lines, the researchers in charge of this project asked several professional photographers to analyze the photos and try to decide which photo had been edited by a professional or semi-professional or which by the artificial intelligence system. The result of this analysis was that the 40% of photos edited by Google's system were classified as human edited.
If you are interested in knowing much more details about this type of projects carried out by Google, tell you that a web page where we can delight ourselves with a complete gallery of images where we can see the original photo and the edition made by Google's artificial intelligence.
Further information: Cornell University