sigmoid.social is one of the many independent Mastodon servers you can use to participate in the fediverse.
A social space for people researching, working with, or just interested in AI!

Server stats:

612
active users

#ocr

4 posts4 participants0 posts today

HHS' Office for Civil Rights Settles HIPAA Privacy and Security Rule Investigation with Deer Oaks Behavioral Health for $225k and a Corrective Action Plan:

databreaches.net/2025/07/08/hh

This was a ransomware attack in 2023 claimed by LockBit. Deer Oaks was already under investigation for a prior breach and HHS OCR expanded their case.

O Nanonets-OCR-s da Nanonets é um modelo OCR que transforma documentos em markdown estruturado, ideal para LLMs. Recursos incluem reconhecimento de equações LaTeX, descrição inteligente de imagens, deteção de assinaturas e marcas d'água, manipulação de caixas de seleção e extração de tabelas complexas.

📎huggingface.co/nanonets/Nanone

📎github.com/NanoNets/docext

📎idp-leaderboard.org/details/

huggingface.conanonets/Nanonets-OCR-s · Hugging FaceWe’re on a journey to advance and democratize artificial intelligence through open source and open science.

If you haven’t checked in on #IMMARKUS lately (understandable—there’s been a lot going on!)—we’ve added even more transcription service options.

You can now run OCR or full-text transcription with a single click using:

• Anthropic Claude
• Azure Computer Vision
• Google Gemini
• Google Vision OCR
• LLaMA & Qwen via kluster.ai
• OCR.space
• OpenAI GPT
• Volcano Engine Doubao 1.5 Vision Pro

Try it out here: immarkus.xmarkus.org

I have a genuine AI problem which I can't find an easy solution for.

We use Google's Vision API to OCR inscription from photographs.

Most of the time it is great, but sometimes it includes homographic characters.

For example - openbenches.org/bench/38224

It has misdetected the letters in "ΤΟΝΙΑ" as Greek rather than Latin.

I can't send a language hint, because people upload images from all over the world.

Is there a good (preferably free) OCR which wouldn't make this mistake?

OpenBenchesIn Loving Memory of ΤΟΝΙΑ HENDRIKS 21st November 2023

easy but decent command line tool for OCR from screenshots? want to be like, okay, paste this image into emacs buffer (org-download), automatically call OCR thing and dump text into the buffer as well. #emacs #ocr

In our quest to make #OCR as easy to use as possible, we've added two new services to #IMMARKUS: Google Gemini and OpenAI GPT. (You’ll need to bring your own API keys to use them.)

Both return full-text transcriptions–no bounding boxes–so you’ll need to select a region before running them. Works with local images and any #IIIF source!

Continued thread

OCR of the above hand-written text:

: think
Dear 2046,% 9° :
we ave going Yo Sur ve.
1 you end up hearing this
sory,l just wont foexy im
4 bd " N y)
} Sov Fyn \ (2-4
Uh Aid 'V a
hal.
Nig
Twill fry faking
care of The eart]

The last two lines sum it up...

stole your image of a social media post, yadda yadda…

❝ COLLINS: “Tulsi Gabbard testified in March that the intelligence community said lran wasn't building a nuclear weapon.”

TRUMP: “l don't care what she said. | think they were very close to having one.” ❞

BTW: the #OCR app for #android is an absolute must have:
OCR - f-droid.org/packages/io.github

source:
universeodon.com/@jaykuo/11470