News for hackers. Stuff that matters.
Re:LlamaCloud and LlamaParse by binarymax on Tuesday February 20, 2024 @02:25PM (#39445685)
Cool! Which OCR engine/model do you use?
Re:LlamaCloud and LlamaParse by pierre on Tuesday February 20, 2024 @02:55PM (#39446076)
EasyOCR, may switch to paddleOCR in the future.
Re:LlamaCloud and LlamaParse by vikp on Tuesday February 20, 2024 @03:34PM (#39446544)
You may want to try https://github.com/VikParuchuri/surya (I'm the author). I've only benchmarked against tesseract, but it outperforms it by a lot (benchmarks in repo). Happy to discuss.

You could also try https://github.com/VikParuchuri/marker for general PDF parsing (I'm also the author) - it seems like you're more focused on tables.

Re:LlamaCloud and LlamaParse by raffraffraff on Tuesday February 20, 2024 @06:31PM (#39448382)
How does surya compare to AWS Textract? A previous employer went through a bunch of different OCRs and ended up using Textract because they found it to be the most accurate overall.
Re:LlamaCloud and LlamaParse by vikp on Tuesday February 20, 2024 @10:12PM (#39449744)
I unfortunately haven't had time to benchmark against more than tesseract.
Re:LlamaCloud and LlamaParse by kergonath on Wednesday February 21, 2024 @02:08AM (#39450946)
That’s my experience as well. I am still looking for alternatives, but Textract is now the baseline.
Re:LlamaCloud and LlamaParse by pryelluw on Tuesday February 20, 2024 @07:39PM (#39448847)
Thanks for sharing.
Re:LlamaCloud and LlamaParse by helloericsf on Tuesday February 20, 2024 @07:51PM (#39448928)
Grateful for your insight! Could you explain the reason for the switch? Is there any benchmark data available for sharing?
Re:LlamaCloud and LlamaParse by pierre on Tuesday February 20, 2024 @11:13PM (#39450080)
Performance depend on the language / type of docs. Main reason for contemplating switching is that easyOCR seems to not be maintained anymore (no commit in the repo in last 5 months)
Re:LlamaCloud and LlamaParse by joaquincabezas on Wednesday February 21, 2024 @04:30AM (#39451790)
PaddleOCR works pretty well, how are you planning to integrate it in your workflow? I found huge differences in throughput between python serving and frameworks (i.e. NVIDIA Triton Inference Server).