Мобильная версияI am currently developing an OCR-based card scanner android app using CameraX and Google ML Kit Text Recognitionv2. While the app performs reasonably well with regular printed text, it struggles to accurately recognize numbers on embossed cards, often mistaking them for alphabetic characters.
Here's a brief overview of our setup:
- We're using CameraX for image capture, ensuring clear and focused images.
- Google ML Kit Text Recognition v2 is employed for text recognition.
- The app primarily scans embossed cards such as debit cards and credit cards, where accurate recognition of numbers is crucial.
Despite our efforts to optimize image quality and preprocessing techniques, the ML Kit consistently misinterprets numeric characters as alphabets, leading to errors in card number extraction. We've experimented with various configurations and settings within ML Kit but haven't achieved satisfactory results.
Has anyone encountered similar issues with ML Kit Text Recognition v2, specifically with recognizing numbers on embossed cards? Are there any specific techniques or configurations we can apply to enhance the recognition accuracy for numeric characters in such scenarios?
Any insights or suggestions would be greatly appreciated. Thank you!
Источник: https://stackoverflow.com/questions/781 ... recognizin
