firas snake
abol3z
AI & ML interests
None yet
Recent Activity
liked a dataset about 1 month ago
nvidia/miracl-vision upvoted a paper about 1 month ago
Beyond Semantic Similarity: Rethinking Retrieval for Agentic Search via Direct Corpus Interaction liked a Space about 1 month ago
Tevatron/BrowseComp-PlusOrganizations
None yet
upvoted a paper about 1 month ago
commented on Supercharge your OCR Pipelines with Open Models 8 months ago
@doladoo yes. I tried Paddle, Miner, Marker, OlmOCR, Chandra-OCR, Docling without VL.
Overall for Arabic, VLM approach showed better performance, and the best was OlmOCR.
Note that my documents are mostly scanned text and tables, nothing more.
commented on Supercharge your OCR Pipelines with Open Models 8 months ago
commented on Supercharge your OCR Pipelines with Open Models 8 months ago
If only this came last week! I spent the last week learning about about and benchmarking all these plus extra models, and I wanna point out a correction. OlmOCR isn't an English language only model, in fact, it produced the best results across all VLM and none VLM frameworks on my Arabic language corpus.
upvoted an article 8 months ago
Article
Supercharge your OCR Pipelines with Open Models


- +5
merve, ariG23498, davanstrien, hynky, andito, reach-vb, pcuenq
• • 315upvoted a paper 11 months ago
upvoted a paper about 1 year ago