[superasn]

It's a great program to write when you start learning machine learning, kinda like the next step to OCR.

Since it's hacker news just wondering what would be the best ML program to do it? Also as for datasets I think you can just download a huge ttf library from the net and write a program to render them into png files.

[amelius]

If you want to take it a step further: there's an app ([1]) that allows you to aim your smartphone camera at a sign, and it will identify the font, translate the sign to a different language, and edit the image with the translated text. All in real time.

[1] https://www.youtube.com/watch?v=h2OfQdYrHRs

[XaspR8d]

I'm not intimately familiar with this app, but I've used Google's version quite a few times. I don't think either really identifies the font much more than separating into a few different categories (e.g. serif vs sans). Not to mention, it struggles with more decorative typefaces.

That said, this tech has been around for years and I still think it's magic every time.

[zeratax]

checking the description, I think it is now google translate.

[fractallyte]

Word Lens was actually introduced on HN a few years back: https://news.ycombinator.com/item?id=2014555