3 ms·

Does anyone have any resources for 'Build it Yourself' form of OCR software. I am aware of Python packages like pytesseract that are somewhat customizable, and

• kumartanmay 7y ago
  It depends from one form to other and sensitivity of data. Our company has developed a product to transform data from pdfs (electronic & non-machine readable files) for financial services. To meet the needs of customers, we have developed parsers for bank statements, loan/lease agreements, valuation reports and portfolio analysis. There are 20% of the documents that are scanned and require OCRs to be trained for different fonts. We wrote about challenges with currents OCRs: https://medium.com/zodhana/modern-day-challenges-with-document-image-analysis-41ffb4a7a956 https://medium.com/zodhana/modern-day-challenges-with-docume...
  • kumartanmay 7y ago
    Usually, OCRs are the goto solution for scanned images, here's why it is not good for extracting financial data with precision: https://medium.com/zodhana/why-ocr-ing-a-bank-statement-is-a-bad-idea-6f976172b7a9 https://medium.com/zodhana/why-ocr-ing-a-bank-statement-is-a...
• eastendguy 7y ago
  Tesseract is the only open-source OCR package that I am aware of. If anyone knows anything else, I would be very curious as well.