[ismailmaj]

I don't know why people mess with tesseract in 2026, attention-based OCRs (and more recently VLMs) outperformed any LSTM-based approach since at least 2020.

My guess is that it's the entry-point to OCR and the internet is flooded by that, just like pandas for data processing.

[mettamage]

Painful comparison haha

Leaving a comment so I can more easily find this

And for the people wondering about Pandas, use Polars instead

[eichin]

I was surprised to learn (from this article) that there are local models that can do this (not sure if there are any that run on hardware I actually have though, unlike Tesseract which works fine on the scanning hardware I set up for it ~5 years ago.) For privacy reasons, cloud-based OCR is a non-starter...

[segmondy]

surprisingly, the ocr models don't need much vram, they are often about 2b, so most 6gb GPU will handle it fine.

[petercooper]

Quite, I threw a so-so photo of an old, long receipt at Qwen 3.5 0.8MB (runs in <2GB) and it nailed spitting 20+ items out in under a second. AI is good at many things, but picking modern dependencies not so much.

[andai]

Are you running it with Ollama?

[petercooper]

LM Studio in this case

[segmondy]

yup, deepseek-ocr-2 will have crushed this. then there's glm-ocr, dots-ocr, etc, paddle-ocr-vl, etc

tons of options ...