Why Is Table Extraction with VLM Models Still Challenging? [D]

About this article

Hey everyone, I’m struggling to find a good approach for converting PDFs to Markdown (especially for financial data). The main challenge is handling borderless tables and tables with more than 5–6 columns. I’ve tried docling, graphite-docling, marker, etc., but haven’t found a solid open-source solution. The only thing that works well so far is LandingAI (but it’s paid). Does anyone know of a good open-source alternative? TIA! Sample: https://preview.redd.it/tajjcvjt5jyg1.png?width=959&fo...