This is no proof that it doesn't send any data when used for conversions online.

It even could work offline, cache/store data and send it when back online.

Looks promising—thanks for sharing! Is it possible to include non-Python binaries somehow? Similar to how it is done with PyInstaller with the --add-binary argument?

P.S.: It has already been mentioned, but I also find the "cowsay" example more confusing than informative. "cowsay" isn't even written in Python, but in Perl, isn't it? Maybe a simple hello-world.py example would work better :)

Thanks!

1. I view that as out of scope but you can always embed your own distribution https://ofek.dev/pyapp/latest/config/#distribution-embedding

2. https://pypi.org/project/cowsay/

As others mentioned, Tesseract is SOTA in FOSS OCR. It also still is being developed, improving slow but constantly.

The main issue for a use-case like NormCap are the trained models: they are optimized for images of _printed_ text and layouts, which is different from on-screen-text in many aspects. Unfortunately, I don't have the resources to train my own models.

Cuneiform was a long time competitor, but afaik development there is stalled.

Author here, excited to find my tool on hn! Happy to answer any questions.

PS: People looking for (FOSS) alternatives, look here: https://github.com/dynobo/normcap#similar-open-source-tools

Here's a list with more tools like this, in case you are interested: https://github.com/dynobo/normcap#similar-open-source-tools

I agree, there are way better cloud based and proprietary OCR solutions out there. But Tesseract still seems to deliver the best results among the FOSS tools, doesn't it?

Back in the days, Cuneiform got close to Tesseract's performance, but AFAIK it wasn't developed further...

Does anyone else know other promising open-source OCR engines?

I didn't know about the story yet, but the reality was already very close to that terrible scenario: https://en.wikipedia.org/wiki/Bachem_Ba_349 (One of the launch sites is quite close to where I live)

That looks like the pilot was intended to survive.

A better comparison would be the Japanese approach. https://en.wikipedia.org/wiki/Kamikaze

More specifically, the Ohka: https://en.wikipedia.org/wiki/Yokosuka_MXY-7_Ohka

Dark Mode is much better for writing code with good quality. As everyone knows: Light attracts bugs!

(SCNR)

Would love to see statistics from the other perspective: Which animals are killed most by humans? Besides insects, I would guess fish and chicken are quite high...

Does this include dust mites in the vacuum cleaner or washing machine?

you mean farmed animals or wild ones ?

In that case, depending on the framing, you might care more about the balance of animals killed by humans vs. animals caused-to-exist by humans. (Or you might think both of those things are bad. Negative utilitarians don’t much like chicken farming, given the large number of chicken lives and the large amount of suffering in each life.)

I see a lot of statements like this (also here in the comments), but is there actual any evidence behind those? Any research results, which has been replicated?

If yes, it would be nice to see it together with such statements, to back them, because I have the feeling most of this is just anecdotical or ideas introduced by some kind of start-up gurus...

Yeah, I agree that these statements are not awfully scientific. I wish I had studies or references to back them up, but I’m not aware of something like that.

However, as others have pointed out, this was originally observed by pg and has become a staple of YC’s advice since then. I think it would be uncharitable to say this is just a hypothetical invented by startup gurus. YC has had a lot of startups pass through their doors, and more experience than almost anyone seeing why people succeed or fail. Their findings may not exactly be science, but I think their advice is as close as we can get right now.