This mobile application prototype aims to help visually impaired people move more independently in urban environments. It detects traffic lights and surrounding objects from a live camera feed and communicates the results through spoken feedback.
Architecture
Android and Kotlin manage the camera feed, touch gestures and audio feedback.
TensorFlow Lite runs an SSD MobileNetV1 model for real-time object detection.
A CNN-based TFLite model classifies red and green traffic lights.
OpenCV supports image preprocessing.
Accessibility
The interface supports audio-led interaction and simple touch controls. On-device inference keeps processing local and reduces latency.
Technologies
Kotlin, Android, TensorFlow Lite, SSD MobileNetV1, CNN and OpenCV.
