Eyesight is not a reason to refuse to work at the computer. First, so you can be absolutely sure that the eyes you will not spoil. And, secondly, the support tools built in to modern operating systems will allow you to easily navigate in unfamiliar applications. But if the sound of the names of the interface elements and text, in General, — it is everyday function for both mobile and desktop devices, what about descriptions of the media files found on the Internet, knew almost no one. But Google found how to solve this problem.
Google is working on integrating Chrome in support of machine learning to generate descriptions of elements that may appear to the user. It was decided to start with images and photos that make up a significant portion of total Internet content. But since they are usually depicted not a particular object but is present and its background, which is important for accurate perception to do with the technology of Google Lens, which determines the names of objects of the real world, it was impossible.
How neural nets work
Judging by the fact that Chrome can now, the recognition is fairly accurate, though not instantly. For example, the photo browser is described as “Fruits or vegetables on the counter.”
It is clear that vegetables in the picture there, and coconuts technically still are nuts, but, first, the algorithms left himself room for maneuver by using the conjunction “or”, and, secondly, I realized that before them is a market counter, which is pretty cool, although fruit it is virtually invisible.
See also: Black screen in Google Chrome? How to fix
Not wrong Chrome here. The browser could easily see not only what is in front of him on the Ferris wheel, but also determined the time of day or night.
Perhaps it would be wise to clarify what the wheels are actually two, and the scene is a Park, but let’s not quibble. In the end, while this is a test version of the technology, which in the future will only improve by teaching myself.
What can Google Chrome
As for the definition of the responsible neural networks, embed them in Chrome in the literal sense would be difficult, if not impossible. So the developers solved the problem by incorporating a mechanism that, after determining the presence of the screen image, it sends the request to Google, which form a description, and then send it to the browser. Hence a delay of several seconds, which may increase depending on the speed of the Internet connection, as well as the complexity of the image. This guarantees that the recognition happens at all, no, because the algorithms are yet unable to formulate a description of just a few million combinations.
Subscribe to our chat in Telegram. Only then will you be able to communicate with like-minded people in a welcoming environment.
As for inclusiveness, then Google is still very far from Apple, which has no equal in this area. The company from Cupertino is one of the first engaged in the introduction of the accessibility features that facilitate the use of people with disabilities. As a result, to manage the Mac computers don’t even have to touch the keyboard or mouse, controlling the events on screen, by voice or auxiliary tools. However, attempts by Google to do something useful for the development of the concept of universal access to the hottest deserves praise.