Scan pdf and read tables with OpenCV & Tesseract OCR

Project Mission: Search and find all tables in a PDF Convert images of tables from PDF (or other) to CSV-formatted tables. Mainly pdf but must use OCR because not all PDF are formatted for parsing Must be able to handle tables that are on two pages (see standard bank report pg 12/13) Requirements: OpenCV (Python) Tesseract v4 A set images of pdfs will be provided. It's important not to optimize the solution for these specific tables. The solution must be generic and will be tested against other images of tables. It is a priority to handle regular tables with high precision. Pie-charts and similar diagrams are a bonus. Proposed steps: 1. Analyze images using OpenCV to determine table cells (rows and columns). 2. Slice input image into multiple images based on cells. 2. Use Tesseract 4 to OCR text from each cell. 4. Output data to CSV Expected outcome: - Conversion is at least 95% accurate with our test-set. Standard tables but not provided to avoid overfitting. - Docker image with all dependencies provided. - Function / Script / API that takes an image and outputs CSV-table. Readings / Links: Improving quality: [login to view URL] Finding text blocks in an image using OpenCV: [login to view URL] Table Analysis using with histogram: [login to view URL] Docker OpenCV Image: [login to view URL] Attached files: pdfs to convert