BankReport Reader - OCR Tool For Bank Reports
summary
BankReport Reader is an OCR pipeline that converts single‑page PDFs to images with pdf2image, converts to NumPy arrays, applies OpenCV preprocessing (resize, grayscale, Gaussian blur, threshold), extracts text with pytesseract, cleans with regex and maps fields by hardcoded coordinates per bank (example: unicredit bank). Backend uses Flask on port 5000, is containerized with Docker on Render and returns JSON; deployment shows HTTP 200 and unit tests (3 unit tests passed) confirm extraction works without any deep‑learning models.
AI-written summary · about
chapters (19)
- 0:00BankReport Reader intro
- 1:07Uploading and processing PDF documents
- 3:40PDF to image conversion
- 5:34Image pre-processing
- 8:01Data extraction to JSON
- 11:02Field coordinates setup
- 14:18Extract data function
- 18:39Text cleaning with regex
- 21:35Handling bank formats
- 26:00Scaling OCR
- 29:18Data mapping and errors
- 30:21Requirements and setup
- 32:32Docker containerization
- 38:01Render deployment
- 42:00Testing with Postman
- 44:42Running the deployed application demo
- 49:24Project results
- 50:58Potential upgrades
- 55:40Q&A
related talks
21mCutting Edge Technologies in Automated Testing
22mOCR, Recognition and Information Extraction from Images
23mText analytics on annual reports
27mHelping SMEs to Get 100% Loan Guarantees through OCR API during Covid
20mScalable OCR pipelines using Python,Tensorflow & Tesseract
26mAdvanced Document Parsing for AI Workflows
19mDocument Reader using Deep Learning with a Small Dataset
01h 00mComplex Data Ingestion with Open Source AI