Internship project
OCR and Document Extraction
OCR and object detection for pulling data out of medical records and bills.
The context
My first professional work: turning paper medical records and bills into structured data that software can use.
What I built
- Compared EasyOCR, PaddleOCR and Tesseract for accuracy on real documents
- Used EAST text detection to find text regions before reading them
- Built a YOLOv8 model to locate fields in health records
- Extracted data from bills
- Prepared a 10,000-image labelled dataset
- Wrote tooling to convert annotations to YOLO format