Apple's Vision framework read the image in 0.27 seconds. A local 27B vision model took 82.8 seconds. Vision made four times as many character errors — but that wasn't what decided it. Vision shredded the table into columns, and a shredded table is syntactically perfect, so nothing downstream can tell it went wrong.
Originally published on hexisteme notes.
I had a 27B vision model running locally (IQ4_XS quantized, 15GB resident) and needed to decide
whether it was worth using for OCR. The comparison was macOS's built-in Vision framework
(VNRecognizeTextRequest) — a dedicated text-recognition engine, free, zero memory
footprint.
Installing OCR, three model tiers side by side, the reflex is "pick the biggest, it's most accurate"....
I ran 110 vision-extraction jobs against a synthetic invoice. The low-resolution modes of the newest...
I benchmarked a vision-language model and scored it at 0.31. The real number was 0.70. Same model,...
The FineBooks project from Hugging Face and EleutherAI tested 14 open-source OCR models on more than 2,000 historical book pages.…
Baidu's Unlimited OCR reads dozens of document pages in a single pass, where previous systems topped out at about ten. A modified…
The Problem That Forced the Question I was building an ingestion pipeline for a personal...
