Braille OCR

To a Braille reader, a page of raised dots is ordinary text. To someone who has never learned the system, the same page can be completely opaque—even when it encodes a language they speak fluently. Learning a new writing system takes time, and manually transcribing an occasional document is rarely practical. Braille OCR bridges that gap by turning photographs of English and Cantonese Braille into readable print text.

Check out the project on GitHub. I also did a demo at a AI Tinkerers Hong Kong meetup, you can find it in the newsletter here: https://post-training.aitinkerers.org/p/top-ai-demos-34-on-device-llms-agent-lab-reports-and-store-automation