←back to thread

DeepSeek OCR

(github.com)
990 points pierre | 5 comments | 20 Oct 25 06:26 UTC | HN request time: 0.956s | source
Show context
farseer ◴[20 Oct 25 06:44 UTC] No.45640648[source]▶
How good is this compared to most commercial OCR software?
replies(1): >>45640663 #
1. ozim ◴[20 Oct 25 06:47 UTC] No.45640663[source]▶
Any vision model is better than commercial OCR software.
replies(3): >>45640798 #>>45640878 #>>45644338 #
2. Etheryte ◴[20 Oct 25 07:31 UTC] No.45640878[source]▶
I'm not really sure if that's an accurate summary of the state of the art, [0] is a better overview. In short, SOTA multi-modal LLMs are the best option for handwriting, nearly anything is good at printed text, for printed media, specialty models from hyperscalers are slightly better than multi-modal LLMs.

[0] https://research.aimultiple.com/ocr-accuracy/

replies(1): >>45641053 #
3. ozim ◴[20 Oct 25 08:00 UTC] No.45641053[source]▶
I see it confirms what I wrote state of art is “not using tessaract anymore” and I think bunch of commercial solutions are stuck with tessaract.
replies(1): >>45641452 #
4. ares623 ◴[20 Oct 25 08:50 UTC] No.45641452{3}[source]▶
I assume Tesseract has the advantage of being able to give a confidence score?
5. dragonwriter ◴[20 Oct 25 14:31 UTC] No.45644338[source]▶
Since “commercial OCR software” includes VLM-based commercial offerings, that's clearly not correct.