A new visual search engine can superimpose data to identify unique points of interest and highlight them in augmented reality.
Israel start-up White Raven is developing a visual search engine for places. The landmark recognition engine uses a camera to extract visual clues from the environment. Consequently, the system can identify places and businesses in the surrounding area. White Raven claims that their deep-learning software can identify landmarks even under major variations in angle, lighting and resolution. The company calls its product, “a middleware between mobile operating systems and location based applications.”
White Raven’s system is designed to be used with a video stream from a moving camera. This makes it potentially useful for helping self-driving cars to find their way. It will also support systems that can superimpose names and information about shops and other places onto a screen as drivers pass by them. White Raven is using public mapping data to create a city-by-city index with information such as addresses and tourist information. The company claims that the system has already achieved a greater than 93 percent accuracy in identifying places. In progression, they have recently begun working with automakers to develop ‘infotainment’ systems for cars. This can not only tell drivers and passengers where they are, but will allow for the creation of augmented reality experiences and games on any connected device with a camera.
We have recently seen deep learning and AI developed for a wide variety of innovative uses, including automating retail and interior design. White Raven has already developed a system for public transport that uses screens in buses, taxis, and trains to superimpose real-time location information on video feeds from the vehicles’ front and rear cameras. It’s also building location-based AR gaming services to keep passengers entertained on their journey. What other uses might there be for very accurate localising software?