Novosibirsk State Technical University has developed an artificial intelligence navigator for obstacle recognition, which will help blind and visually impaired people navigate the terrain.
An inter-faculty development team from the Faculty of Radio Engineering and Electronics and the Faculty of Automation and Computer Engineering of NSTU-NETI has created a video navigator for the visually impaired based on artificial intelligence, which is characterized by efficiency and simplicity. According to the developers, all that is required is an Android phone.
"The person downloads the program we developed, turns on the camera and attaches the gadget to his chest. Then a specially trained neural network works. From the video stream of frames, AI recognizes the situation, the distance to potentially dangerous objects and notifies the user about it with sound," said Yegor Avdeev, a third—year student at AVTF.
To recognize objects, developers use a convolutional neural network model that is trained on manually labeled data. AI can warn about objects at a distance of about five meters, so as not to overload the user with information, as well as respond to unexpected obstacles such as cars or closed gates — the application processes frames in 300 ms.
The development of Novosibirsk university students differs from domestic and foreign analogues in that it is able to work offline and does not contain advertising and a paid subscription. In addition, the program will be able to offer its objects for recognition, so that you can ensure maximum comfort when moving.
The team is currently working on the main mode of operation of this application. "In three months, we will try to release a version that can help with obstacle detection and orientation. There are also ideas for adding other functions to the application, for example, building a route based on a voice request for an address or searching for objects by name," said Egor Avdeev.
The project, according to the developers, is almost ready. Before the release of the new version, the team will conduct testing and further plans to fix minor issues online that do not affect the operation of the main functionality of the application.
The initiator of the project was Andrey Nikulin, Associate Professor of the Department of Radio Receivers and Radio Transmitting Devices at the Faculty of Radio Engineering and Electronics, who last year developed a hand assistant for blind and visually impaired people.
The project of the team with Automation and Computer Engineering Faculty has already been highly appreciated by the jury in the senior League student project competition and won first place among IT projects.