Students in Siberia Build Offline AI Guide for the Blind

A new Android-based wearable helps visually impaired users navigate city streets safely.

Students from Novosibirsk State Technical University have developed an AI-powered video navigation app that helps blind and visually impaired individuals move around cities with greater safety. All it takes is an Android smartphone and a chest-mounted camera.

A neural network processes real-time video to identify obstacles and alert users with audio cues. The system uses a convolutional neural network trained on labeled datasets. It can detect objects up to five meters away and reacts quickly to sudden obstacles, such as cars or closing gates.

The app processes visual data in just 300 milliseconds, helping prevent collisions. It works without an internet connection, requires no subscription, and is compatible with entry-level smartphones. Future updates aim to include voice commands, object search by name, and the ability to train the AI on new objects. The project has already taken first place in a national student IT competition. Final testing is scheduled before public release.