When will OCRopus work as well as current commercial OCR systems?
We hope to achieve commercial error rates (but not commercial throughput) on the kinds of documents we are interested in (books, journal articles, etc.) by the beta release; please see the roadmap on the home page. Other document types (letters, bills, magazines, etc.) may work reasonably well, but if you’re specifically interested in them, please consider contributing your time to the project to help improve performance on those document types.