[sushid]

Making OCR more accurate for regular text (e.g. data extraction from documents) would be useful; not sure how useful code transcription is