4 ms·

OCR from scratch is pretty difficult, believe me, I have tried and even with OpenCV support you would need months or years (don't know if openCV is shipped with

OCR from scratch is pretty difficult, believe me, I have tried and even with OpenCV support you would need months or years (don't know if openCV is shipped with a OCR solution these days). Even if you restrict yourself to a small subset of characters and fonts you need to solve a lot of problems. There are really good ready to use solutions though. Tried the cloud service from ABBYY which nets really good results, but I guess you wouldn't want that. The project is an interesting idea though. Would save time in writing all those protocols... To answer the question, I think being paranoid about such things is very healthy.

ge96 5y ago
Yeah I don't have a vision processing background so it would be brutal if I tried from scratch. It's just that idea that someone could be peaking into your "soul" ha, if they can screenshot your screen/which can display whatever. It would definitely need to be a local program. Edit: another thing is I want to use a Tensor Flow JS model and want to make sure whatever is passed into it isn't also getting sent out somewhere.