Design And Implementation of an Automated OMR Analyser Using ML and Image Processing

Abstract

Automated evaluation of OMR (Optical Mark Recognition) answer sheets plays a significant role in academic and competitive examinations. Traditional OMR systems depend on expensive scanners and strict sheet templates, leading to operational limitations and high processing costs. This work presents a web-based OMR evaluation system using Machine Learning and Image Processing techniques to accurately detect filled bubbles from scanned or photographed sheets. The system integrates OpenCV for preprocessing and a Convolutional Neural Network (CNN) model for classification of filled and unfilled bubbles. The solution supports custom OMR formats, handles variations in lighting, shading, and sheet alignment, and delivers fast and accurate results with real-time visualization. Experimental results show detection accuracy of above 98%, demonstrating its effectiveness for large-scale examination environments..

Introduction

The study presents an Automated OMR Analyzer designed to evaluate multiple-choice answer sheets without specialized hardware, addressing the limitations of traditional OMR systems, such as cost, inflexibility, and manual errors. The system combines image processing and machine learning, using techniques like grayscale conversion, Gaussian blurring, adaptive thresholding, contour detection, and perspective correction to preprocess scanned or camera-captured OMR sheets. Region of Interest (ROI) extraction identifies bubble positions, and a Convolutional Neural Network (CNN) classifies each bubble as filled or unfilled, handling faint, partially marked, or smudged responses.

Implemented as a web-based platform, the system allows users to upload templates, filled sheets, and answer keys, producing instant scoring, visual analytics, and downloadable reports. The architecture supports flexible OMR formats, bulk evaluation, and concurrent users. Testing demonstrated high accuracy (~99.2%), robustness against variations in lighting, alignment, and marking style, and faster, error-free evaluation compared to traditional pixel-counting methods. The solution is cost-effective, scalable, and adaptable, making it suitable for educational institutes and examination authorities.

Conclusion

The proposed Automated OMR Evaluation System successfully demonstrates an efficient, accurate, and scalable solution for processing and assessing OMR answer sheets using Machine Learning and Image Processing techniques. By integrating OpenCV-based preprocessing with a CNN-based bubble classification model, the system effectively identifies filled and unfilled responses under varying lighting conditions, marking intensities, orientations, and image qualities. Unlike traditional hardware-dependent OMR scanners, the developed web-based platform provides flexibility by supporting customizable OMR templates and eliminating the need for expensive scanning equipment, enabling evaluation through scanned or mobile-captured images.Experimental results validate that the system achieves high accuracy, rapid processing time, and strong robustness, making it suitable for real-time exam evaluation environments. The automated scoring, result visualization, and downloadable reports significantly reduce human workload while improving transparency, consistency, and reliability.Future work includes extending the system for bulk OMR sheet processing, enhancing multi-class bubble classification, deploying cloud support, and developing fully mobile-based real-time evaluation

References

[1] Zeki Kucukkara, Abdullah Erdal Tumer, “Image Processing Oriented Optical Mark Recognition and Evaluation System”, Journal of Image Processing and Pattern Recognition, Vol. 12, 2018, pp. 68-75. [2] Pooja Raundale, Taruna Sharma, Saurabh Jadhav, Rajan Margaye, “Optical Mark Recognition using OpenCV”, International Journal of Computer Applications, Vol. 152, 2019, pp. 45-52. [3] Harendra Kumar, Himanshu Chauhan, Sonam Mittal, “Analysis of OMR Sheet Using Machine Learning Model”, International Journal of Advanced Research in Computer Science, Vol. 10, 2019, pp. 84-91. [4] Qamar Hafeez, Waqar Aslam, M. Ikramullah Lali, Shafiq Ahmad, Mejdal Alqahtani, Muhammad Shafiq, “Fault Tolerant Optical Mark Recognition”, Proc. of IEEE, Vol. 45, 2023, pp. 321-330. [5] Effat Somaiya, Alifa Sun Mim, Mohammed Abdul Kader, “Webcam-Based Robust and Affordable Optical Mark Recognition System for Teachers”, Journal of Image and Video Processing, Vol. 36, 2023, pp. 100-107 [6] Pham Doan Tinh, Ta Quang Minh, “Automated Paper-Based Multiple Choice Scoring Framework Using Fast Object Detection Algorithm”, Proc. of International Conference on Computer Vision, Vol. 40, 2024, pp. 250-257. [7] Rusul Hussein Hasan, Inaam Salman Aboud, Rasha Majid Hassoon, Ali Saif Aldeen Aubaid Khioon, “Optical Mark Recognition using Modify Bi-Directional Associative Memory”, Journal of Computer Vision and Applications, Vol. 42, 2024, pp. 111-120. [8] Sujal Rooge, Veeresh Kotresh Kadli, Shivanand Manjare, Varun Kumar M B, Suguna A, “OMR Detection Using Image Processing Technique”, International Journal of Advanced Computer Science, Vol. 39, 2024, pp. 180-188 [9] Dharmik R. C., Rangari S., Jain S., Nilawar A., Deshmukh G., Yeole A., “Optical Mark Recognition Evaluation System using Dual-Component Approach”, International Journal of Data Analysis, Vol. 21, 2024, pp. 203-210. [10] Rushikesh G. Dongare, Prof. Tanuja S. Mulla, Gauri Y. Kakulte, Manjiri M. Tamkar, Shraddha S. Tambekar, “OMR System Empowered by Machine Learning and Image Processing”, International Journal of Computer Applications, Vol. 52, 2024, pp. 142-1.

Copyright

Copyright © 2025 Ravindra S, Ranganath K N, Nithin A, Manoj M H, M S Hemanth Kumar. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.