Google has done it again! It seems that the tech giant has plenty of innovations hidden up their sleeves and this time, they don’t even need internet connection. Artificial intelligence is allowing a simple picture the ability to identify landmarks, objects and faces – at times, they can even do better than how humans perceive things.
Just this June, Google announced MobileNets software that will open up a neural network of computational imaging. They designed these for other programmers to integrate into their apps, so it can specifically operate on mobile devices’ smaller hardware. This will help conquer a few of the biggest difficulties in delivering computer imaging to mobile phones with a design that makes the most of the capabilities of mobile processors.
The program doesn’t just form new abilities, but it delivers computational imaging into a small package that will operate using a smartphone without collecting information on a cloud. This means that apps use the programming without needing internet. The programming provides tablets and smartphones the capacity to identify people and objects, as well as, famous landmarks. Google even records fine-grain arrangements as a part of its potential uses, like recognizing the breed of a certain dog.
The release of the program for mobile users only means that third-party apps could soon acquire improved or new computational imaging features. By making an open source programming, Google is unveiling the software for use beyond just Google-owned apps. They can expand the app for various users from augmented reality to reverse image searches.
The capacity to identify faces and objects in a picture through a neural network is not a new method, but MobilNets are more competent. They are developing a faster, smaller program to be used on smartphones, even without internet.
According to Menglong Zhu and Andrew Howard, Google software engineers, deep learning has prompted remarkable developments in the sphere of computer vision recently, with neural networks frequently leading the frontier of visual recognition technology.
Even though several of these technologies, like text, logo, object and landmark recognition are already available for devices connected using the Cloud Vision API, they trust that the constantly growing computational capabilities of smartphones can allow the distribution of these technologies to users anywhere and anytime with or without internet connection.