BrailleLens: Real-Time Fingertip-Guided Braille Recognition for Sinhala

Access to Braille instruction remains limited in Sri Lanka and other low-resource settings, where qualified tutors are scarce and optical Braille systems largely focus on page-level transcription. We present BrailleLens, an interactive mobile app for Sinhala Braille learning, verified on both a standard phone camera and AiSee smart glasses. It combines YOLO26-based cell detection, CNN classification, deterministic Sinhala Braille decoding, and fingertip-guided interaction to provide cell-level spoken feedback and assessment. The fully on-device Flutter pipeline supports offline Sinhala speech. The cell classifier reaches over 99% accuracy on public Braille datasets and 95.93% after Gold fine-tuning, while the full pipeline reaches 83.4% end-to-end page accuracy. BrailleLens is currently a pilot-stage system evaluated by the development team, with evaluation involving Braille readers planned as future work.