thiagolima

4 ms·

1.
by thiagolima 1mo ago
Leo, i benchmarked Qwen2.5-VL-3B, 4-bit via MLX, against Tesseract on the same 24 samples: https://thiagotigaz.github.io/ocr-it/bench/ Its much lower and the error rate is much higher. It works, but for our usecas
2.
by thiagolima 1mo ago
Hi guys, the chrome extension was finally approved and is now available at https://chromewebstore.google.com/detail/ocr-it/pekiamimgjbj... We are still waiting for the firefox extension to be approved, i will pos
3.
by thiagolima 1mo ago
We did some benchmarking against other libraries like RapidOCR and EasyOCR results in https://thiagotigaz.github.io/ocr-it/bench/
4.
by thiagolima 1mo ago
I added Firefox support. 0.3.0 builds for both browsers from the same source, and it's submitted for both google/chrome and mozilla/firefox, i am waiting on reviews now, which usually takes a few days. Until it's approve
5.
by thiagolima 1mo ago
I did a small benchmark for RapidOCR: https://thiagotigaz.github.io/ocr-it/bench/ For the input text rendered on screen, Tesseract did better on both accuracy and speed. We got about 0.1% character error vs 1–2% f
6.
by thiagolima 1mo ago
it can also auto paginate for you, no need to keep hitting the hotkey every page. It can paginate by hotkey, xy point on screen or selector.
7.
by thiagolima 1mo ago
Haven't tested rough scans honestly — my own use is rendered text, the easy case, where it's 93-95% confidence. I've been feeding it Kindle trading books into my finance app to compare strategies against my codebase, and an L
8.
by thiagolima 1mo ago
Half the context I want to give a model is locked inside something I can't select from: a scanned book, a slide deck, a course viewer, a "PDF" that's really page images. Copy-paste gets you nothing, and screenshotting 20
9.
OCR It – pull text out of un-copyable documents for your LLM(github.com)
140 points by thiagolima 1mo ago|35 comments
10.
Mergecrew: Open-source agentic SDLC with human-gated prod deploys(github.com)
2 points by thiagolima 5mo ago|0 comments