We might all marvel at the advancements in computer vision and artificial intelligence, but most of the applications we applaud involve us seeing their results or even their targets. A group of researchers inside Microsoft are working on an implementation of these technologies not just to let computers see but to also let humans see, specifically those that have lost their normal vision. Called Seeing AI, the research project is an attempt to combine years, even decades, of computer vision, machine learning, and artificial intelligence research into an all-in-one app for the visually impaired.
There are already existing apps for the blind. In fact, there might be a good number of them already. The problem with almost all of them, however, is that they are single purpose apps. One can be used for reading text and nothing else. Another can, perhaps be used for location and that’s that. Switching between them as the context arises isn’t just inconvenient, it’s also impractical.
Seeing AI then tries to be the be all and end all of such apps. It tries to do everything to properly identify what a person is looking at, from text, people, things, and even emotions. Yes, it can try to gauge and identify whether a person is sad, happy or surprised. It even tries to guess the person’s age.
Sounds familiar? Microsoft has actually launched research experiments like that in the past, trying to guess people’s ages based on their web cam snapshot. Ultimately, Seeing AI’s magical abilities is a fruit of Microsoft’s research in the field of computer vision, mixed with advancements in machine learning, neural networks, natural language processing, and artificial intelligence. Part of that magic is available for everyone to try out in the newly launched Caption Bot website that takes any picture and tries to identify the contents.
The Seeing AI will be available as a smartphone app but also as something that can be used on the Pivothead smart glasses. No launch date, however, has been given yet.