emirates7 - Students at Novosibirsk State Technical University have created an AI-powered video navigation app aimed at improving urban mobility and safety for blind and visually impaired users.
The system operates using just an Android smartphone paired with a chest-mounted camera.
Utilising a convolutional neural network, the app processes real-time video footage to detect obstacles and deliver audio warnings. Trained on labelled datasets, the AI can identify objects up to five metres away and reacts quickly to sudden dangers like moving vehicles or closing gates.
The visual data is processed within 300 milliseconds, significantly reducing the risk of collisions. The application functions offline, requires no subscription, and is compatible with basic smartphones.
Future updates are set to introduce voice command functionality, object identification by name, and user-driven AI training to recognise new objects. The project recently earned first place in a national student IT competition, with final testing underway before its public launch.
The system operates using just an Android smartphone paired with a chest-mounted camera.
Utilising a convolutional neural network, the app processes real-time video footage to detect obstacles and deliver audio warnings. Trained on labelled datasets, the AI can identify objects up to five metres away and reacts quickly to sudden dangers like moving vehicles or closing gates.
The visual data is processed within 300 milliseconds, significantly reducing the risk of collisions. The application functions offline, requires no subscription, and is compatible with basic smartphones.
Future updates are set to introduce voice command functionality, object identification by name, and user-driven AI training to recognise new objects. The project recently earned first place in a national student IT competition, with final testing underway before its public launch.