TLUG Mailing List

> Quite closely related, I've been wondering what the state of the art for
open-source OCR is, particularly of Japanese text.

This is going to be one of those "give me an apple pie recipe", "here is my pumpkin pie one" type of replies, but Google's Cloud Vision API, while not FOSS, is compelling for Japanese OCR.

I have been using this on mid-to-late Showa books and it has been excellent. 1000 images free, and $1.50 per 1000 images thereafter. Assuming full pricing, it comes out to.21 yen per page.

Flow looks like this:

Imagemagick to make jpgs for each pdf page.
Single command to copy into cloud in parallel, publishing to a request queue, triggering OCR via short script.
Small local python script subscribes to the results stream (topic), and downloads/saves results as they are completed.

This post compares Cloud Vision and Tesseract for English, and finds Cloud Vision more accurate.

https://medium.com/ixor/comparing-tesseract-ocr-with-google-vision-ocr-for-text-recognition-in-invoices-bddf98f3f3bd

This from 2023 compares Japanese free OCR packages, finds PaddleOCR (https://github.com/PaddlePaddle/PaddleOCR) best for their use case, but also notes that Cloud Vision dominates.

https://zenn.dev/piment/articles/254dde3ecf7f10

On Mon, Sep 9, 2024 at 12:00 PM <tlug-request@example.com> wrote:

Send Tlug mailing list submissions to
tlug@example.com

To subscribe or unsubscribe via the World Wide Web, visit
https://lists.tlug.jp/mailman/listinfo/tlug
or, via email, send a message with subject or body 'help' to
tlug-request@example.com

You can reach the person managing the list at
tlug-owner@example.com

When replying, please edit your Subject line so it is more specific
than "Re: Contents of Tlug digest..."

Today's Topics:

1. Re: Anyone alive out here ? (Chris Salisbury)
2. Re: [announcement] September 14th Technical meeting
(Edward Middleton)
3. Re: [announcement] September 14th Technical meeting
(Brian Clemens)
4. Re: Anyone alive out here ? (Benjamin Kowarsch)
5. Re: [announcement] September 14th Technical meeting (Raymond Wan)
6. Re: [announcement] September 14th Technical meeting
(Edward Middleton)

----------------------------------------------------------------------

Message: 1
Date: Sun, 8 Sep 2024 12:12:29 +0900
From: Chris Salisbury <chris.salisbury@example.com>
To: Tokyo Linux Users Group <tlug@example.com>
Subject: Re: [tlug] Anyone alive out here ?
Message-ID:
<CAH2XypHG2YUiYrURRX-PPYz9+oN5BKPLmRxvY3X6jvFjJfvZww@example.com>
Content-Type: text/plain; charset="utf-8"

>>> What's the current state of the art? Interfacing models with text has
big limits...
>> Quite closely related, I've been wondering what the state of the art for
open-source OCR is, particularly of Japanese text.
> I'm waiting for the first Llama-like LLM with image recognition similar
to ChatGPT.

Not sure if "similar to ChatGPT" precludes the much worse performance of
the "Llama-like" models at https://ollama.com/search?c=vision, but I've
been impressed with llava-phi3 for a model small enough to run on a phone
that I got for free with a 2-month contract 3 years ago. I do this on
(unrooted) android 11 with termux->proot->ollama. But I've never tried OCR
with it, Japanese or otherwise.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.tlug.jp/ML/attachments/20240908/783669b0/attachment.htm>

------------------------------

Message: 2
Date: Sun, 8 Sep 2024 16:56:21 +0900
From: Edward Middleton <edward.middleton@example.com>
To: tlug@example.com
Subject: Re: [tlug] [announcement] September 14th Technical meeting
Message-ID: <5e3c9ce4-f94b-4016-8284-fd0c28fdd18a@example.com>
Content-Type: text/plain; charset=UTF-8; format=flowed

On 3/9/24 22:12, Joseph Hart wrote:
> Any chance of remote viewing?? Tokyo is a bit far from Kobe...

It should be possible if someone can help. I don't think I will have
time to set that up.

Edward

------------------------------

Message: 3
Date: Sun, 8 Sep 2024 17:29:08 +0900
From: Brian Clemens <brian@example.com>
To: Tokyo Linux Users Group <tlug@example.com>
Subject: Re: [tlug] [announcement] September 14th Technical meeting
Message-ID:
<6tnaimnmg27nsgk32ofezu6wyjpxptdk3m62eaxfjrgo7pr5qb@jsiayxzmqv74>
Content-Type: text/plain; charset="utf-8"

> It should be possible if someone can help. I don't think I will have time
> to set that up.

I think I can arrange that, I've got a wireless lapel mic and whatnot.

--
Brian Clemens
Cofounder and Vice President, Rocky Enterprise Software Foundation
Board Vice Chair, Linux Professional Institute
Open Source Program Administrator, CIQ
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 228 bytes
Desc: not available
URL: <https://lists.tlug.jp/ML/attachments/20240908/73d30a96/attachment.sig>

------------------------------

Message: 4
Date: Sun, 8 Sep 2024 18:05:39 +0900
From: Benjamin Kowarsch <trijezdci@example.com>
To: jfhart085@example.com, Tokyo Linux Users Group <tlug@example.com>
Subject: Re: [tlug] Anyone alive out here ?
Message-ID:
<CADR0rnfiGgmD4=82rstqEeqi9ryR9_80dGYGtzEL0xbSaToOhA@example.com>
Content-Type: text/plain; charset="utf-8"

On Mon, 2 Sept 2024 at 18:57, J. Hart wrote:

> What's everyone working on ?
>

Welding exercises using aluminium alloy wire with my TIG welder so I can
weld together a new mounting bracket for the tow bar of my DIY
bicycle trailer. No software involved there, but welding aluminium is not
for the faint of heart either.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.tlug.jp/ML/attachments/20240908/37afc07b/attachment.htm>

------------------------------

Message: 5
Date: Sun, 8 Sep 2024 11:18:24 +0100
From: Raymond Wan <rwan.kyoto@example.com>
To: Tokyo Linux Users Group <tlug@example.com>
Subject: Re: [tlug] [announcement] September 14th Technical meeting
Message-ID:
<CAAhy3dvq1RePiOi8eeFCHxud3oHGpHB6_v3_yWKRtr11oszLXA@example.com>
Content-Type: text/plain; charset="UTF-8"

On Sun, Sep 8, 2024 at 9:04?AM Edward Middleton
<edward.middleton@example.com> wrote:
> On 3/9/24 22:12, Joseph Hart wrote:
> > Any chance of remote viewing? Tokyo is a bit far from Kobe...
> It should be possible if someone can help. I don't think I will have
> time to set that up.

If it gets recorded in the end, it would be great to get a link. If
you are making slides, a copy of the slides would already be very
helpful.

I have a feeling that I will need to learn more about 3D printing at
some point in the future.

Sorry I couldn't be there. Hope you guys have fun!

Ray

------------------------------

Message: 6
Date: Sun, 8 Sep 2024 19:50:43 +0900
From: Edward Middleton <edward.middleton@example.com>
To: tlug@example.com
Subject: Re: [tlug] [announcement] September 14th Technical meeting
Message-ID: <921d86cf-8cae-4498-9012-2bfcf523a0c6@example.com>
Content-Type: text/plain; charset=UTF-8; format=flowed

On 8/9/24 19:18, Raymond Wan wrote:
> On Sun, Sep 8, 2024 at 9:04?AM Edward Middleton
> <edward.middleton@example.com> wrote:
>> On 3/9/24 22:12, Joseph Hart wrote:
>>> Any chance of remote viewing? Tokyo is a bit far from Kobe...
>> It should be possible if someone can help. I don't think I will have
>> time to set that up.
>
> If it gets recorded in the end, it would be great to get a link. If
> you are making slides, a copy of the slides would already be very
> helpful.
>
> I have a feeling that I will need to learn more about 3D printing at
> some point in the future.

This is very much a, from nothing to building your own commercial grade
printer kind of talk. It will try to cover everything just not too
deeply. The hope is that by the end I will have either inspired you to
get into DIY 3D printing or warned you off.

I am collecting material on logseq and will probably use presentation
mode on that unless I get a bunch more time and energy before next week.
There is a huge amount of really good material on 3d printing and
specifically voron printers it just takes time pulling out the best bits.

Edward

------------------------------

Subject: Digest Footer

--
To unsubscribe from this mailing list,
please see the instructions at http://lists.tlug.jp/list.html

------------------------------

End of Tlug Digest, Vol 223, Issue 8
************************************

Re: [tlug] Anyone alive out here ? (state-of-the-art OCR)