Talking about fast-growing technologies, at just a couple of months after the beta launch of Google’s Glass, a new app, able to recognize the object the user is looking at, has been developed.
If you are wondering who thought about it first, well, it’s AlchemyAPI, a company which provides language processing services all over the world. The application which can recognize the object an user is looking at, was developed during an intern hack session, where the company brainstormed about new methods of using the image recognition service they recently launched. And it isn’t surprising they thought of using it with the Google Glass, since the device’s camera is able to shoot quite accurate shots.
Before telling you (or displaying) the name of the object you are looking at, the app will take photos using Glass’s camera, or allow the user to take them. Next, the photos are sent to a computer (if in range), or to the cloud in order to process it with the AlchemyAPI recognition service. In less of a second, the software should send back to the Glass the name of the object it sees.
Elliot Turner, the CEO of AlchemyAPI’s says that the software needs around 250 ms to give a verdict.
The application is using for the recognition process a deep learning software based on networks similar to the neuronal connections, able to learn patterns faster and more efficient than traditional systems. Google was the first to develop such a system and as far as we know Microsoft, and now Facebook are also investing in developing and growing their own. On the other hand, one could assume that, regardless to how smart the system is, it doesn’t do much and having an app which doesn’t have a very practical use is just another straw to stack. However, it is remarkable that a computer can now do all those actions in such a short time frame.
Even though the software can be improved, its overall performance is quite astonishing. If in the near future, the company could improve its competency and increase its capability of giving precise information, it could be quite the catch. The company’s CEO said that some of the first companies to be interested in the image recognition service are companies which handle a lot of media content and need it to label and categorize this type of content.
Another one of these object recognition systems developed by AlchemAPI is used in the Google image search engine of the Google+. Turner claims that his company’s recognition system has a fail average of only 17 percent, which is around the same percentage the other recognition systems on the market achieve. By the way, if you want to test some of these recognition systems, you can prove it against the ImageNet database, one of the biggest media content of the world containing around 50 million images in 20.000 categories.