Hacker News Reader

Comment by melissalobos | Hacker News Reader

Thanks for sharing, this looks really useful. I have started biking again recently and the labels on the tire are super confusing. Do you have any plans to extend this(attempt OCR or something)?

I tried looking at the source a bit and it seems like your markup is pretty clean, if a bit unreadable after minification. I liked the classes "question" and "answer". It looks like you are planning on using affiliate links to monetize it?

View on HN

Thank you for the feedback. OCR is a great idea, for me it seems to be quite difficult to implement and probably requires a mobile app.

The affiliate links are just a side effect to pay for the hosting.

You could possibly do it server side. I would personally implement it as a Flask server that took an image, ran it through either Pytesseract (quick, got to manually preprocess, needs tesseract installed) or Easyocr (slower but works out the box) then send back the if and only if it was a valid tire-info-string to avoid people abusing it for other purposes.

Wow, thank you, I will check that.

There is a barcode scanner on https://schnelltesttest.de/

Their source code is here: https://github.com/zerforschung/schnelltesttest.de

Thanks for the tip, I will look into it.

You could use a simple client-side JS OCR library, like GOCR.js[0]. It may only work on good photos of relatively new tires with clearly printed text, and that's not always what bicycle tires look like when you're looking to replace it, so not sure how useful it'd be.

[0]: https://antimatter15.com/gocr.js/demo.html

Interesting approach, thank you. Sure, old tires might be hard to scan, but definitely worth to think about.

I tried to to use GOCR packaged with Ubuntu on playlists, expecting it to work well since it's computer output. But it turned out not working so well. I also tried scaling the text but that didn't fix it.

Can I ask why OCR, other than to learn something new. From a functionality standpoint you’ve hit the nail on the head. Some suggestions in this thread however start to take away from the appeal of the site. I suspect the target audience will much rather type out the dozen numbers/chars than they will attempt to mess around with taking a picture and having OCR take over. I’d be willing to bet the time to enter data is comparable and if there are any savings to be had from OCR (doubtful) it would be measured in seconds at best.

Identify the audience and build the product for it. A tech savvy audience (HN) would love to see OCR because we have nothing better to do than look for “latest cool gadgets”.

Good point, the page is designed for easy usage and beginners, rather than experts.

Reply on news.ycombinator.com