@db_geek gonna be honest, that's pretty underwhelming 馃珷
Post
Replies:
10
In return for being a named sponsor the I can just about cover that particular expense 馃榿
@simonzerafa @openbenches
Ultra Gold Platinum Exclusive Sponsors start at 50p 馃榿
OK, I'm in. More than happy to sponsor at that level or up to 拢10 to cover anticipated future expenses 馃檪
@simonzerafa
That's incredibly kind of you 馃グ
We have donation options at https://openbenches.org/support
GitHub Sponsors or OpenCollective
https://github.com/sponsors/openbenches
https://opencollective.com/openbenches
I don't know, what OmniTools is using for OCR in the background.
It is easy to run it locally.
For testing purposes also a demo is available:
https://omnitools.app/image-generic/image-to-text
@db_geek gonna be honest, that's pretty underwhelming 馃珷
Yes, that looks really bad.
@Edent I am currently doing a lot of vision work on documents, PaddleOCR for a fully local setup is doing decently, might be worth looking into.
Compared to Teserract its setup for scene text detection, so it got a chance to work for benches :)
@fallenhitokiri
Oooh! Paddle's API is both generous and accurate.
It's a pity it is so slow. Not sure if that's a latency to China thing, or just the speed of the model.
Will keep trying. Ta!
