News for hackers. Stuff that matters.
Re:Mistral OCR by techwizrd on Thursday March 06, 2025 @03:47PM (#43284920)
The challenge I have is how to get bounding boxes for the OCR, for things like redaction/de-identification.
Re:Mistral OCR by kbyatnal on Thursday March 06, 2025 @04:18PM (#43285166)
yeah that's a fun challenge — what we've seen work well is a system that forces the LLM to generate citations for all extracted data, map that back to the original OCR content, and then generate bounding boxes that way. Tons of edge cases for sure that we've built a suite of heuristics for over time, but overall works really well.
Re:Mistral OCR by dontlikeyoueith on Thursday March 06, 2025 @05:24PM (#43285654)
Why would you do this and not use Textract?
Re:Mistral OCR by schcrosby on Thursday March 06, 2025 @07:04PM (#43286301)
I too have this question.
Re:Mistral OCR by dontlikeyoueith on Thursday March 06, 2025 @05:24PM (#43285651)
AWS Textract works pretty well for this and is much cheaper than running LLMs.
Re:Mistral OCR by daemonologist on Thursday March 06, 2025 @05:48PM (#43285815)
Textract is more expensive than this (for your first 1M pages per month at least) and significantly more than something like Gemini Flash. I agree it works pretty well though - definitely better than any of the open source pure OCR solutions I've tried.