Mark Extraction System

This project presents a custom-built solution to automate the extraction of student marks from scanned answer sheets, specifically designed for use in our college.Using a Transformer-based OCR model trained to recognize marks in our institution's answer sheet format, the system accurately reads handwritten or printed marks assigned to each question.The scanned images are preprocessed using OpenCV (cv2), and deep learning-based text recognition is carried out using PyTorch (torch).This approach ensures high accuracy even with varied handwriting styles commonly seen in manually corrected papers.After extracting the marks, the data is visualized using matplotlib and automatically inserted into a structured Excel file, eliminating the need for manual entry.This not only speeds up the result processing workflow but also reduces the chances of human error.The system is tailored for our college's internal evaluation needs, making it highly practical, scalable, and ready for real-world academic deployment.