A smartphone application using a custom-trained neural network to detect and announce traffic signs, pedestrian crossings, and traffic lights in real time — giving visually impaired people independent, safe navigation of road environments.
For visually impaired pedestrians, navigating traffic environments is a daily challenge. Existing assistive technology focuses mostly on obstacle detection — but traffic signs, crossing lights, and pedestrian zones are equally critical and largely ignored by consumer-grade assistive tech.
K.I.V.E was our diploma project at HTL Braunau — a year-long deep dive into training a custom computer vision model specifically for Austrian traffic signage and pedestrian infrastructure. We collaborated directly with visually impaired testers throughout development to ensure the output was genuinely useful, not just technically impressive.
The system uses the smartphone camera as a sensor, processes frames through a custom-trained object detection model, and delivers spatial audio announcements through the phone speaker or earphones — telling the user not just what was detected, but how far away it is and whether it requires immediate action.
┌────────────────────────────────────────────────────────┐ │ SMARTPHONE APP │ ├────────────┬──────────────────────┬────────────────────┤ │ Camera │ AI Inference Engine │ Audio Output │ │ Module │ ────────────────────│ ──────────────────│ │ live │ Custom model │ TTS speech synth │ │ frames │ Bounding boxes │ Distance announce │ │ → frames │ Confidence score │ Priority filter │ │ │ Class labels │ Speaker / earphone│ └────────────┴──────────────────────┴────────────────────┘ Detection flow: Frame → [Preprocessing] → [Model Inference] → [Distance Calc] → [Priority Filter] → [Audio Announce] Example outputs: "Red traffic light, approximately 4 metres ahead" "Pedestrian crossing detected, 6 metres to your left" "Green pedestrian signal, safe to cross"
All three team members are graduates of HTL Braunau's Elektronik und technische Informatik department, with the autonomous focus area "Communications". The project was completed as the compulsory diploma thesis (Diplomarbeit) in the final year of the five-year programme.