Hacker News new | ask | show | jobs
by glandium 1291 days ago
Oh wow, first time I see an OCR that can deal with 行書体. How extensive is your dataset for 行書 and do you also have 草書 or 隷書?
1 comments

Under the hood, it's actually using Google's Cloud OCR API. Not sure how well it handles 草書 or 隷書, but it does a pretty remarkable job with 行書. Also does surprisingly well with super pixelated text. I can look up words in 16-bit games trying to squeeze kanji into 16x16 sprites or something, and it holds up incredibly well.

You can also swap out the OCR API it uses. Right now I support Google and ocr.space.