Digital Loan Management Solution for Sarvdhan
OMR
Efficient and accurate system to process OMR sheets
Client Overview
The project began by addressing the challenge of multiple exam formats, ranging from 50 to 300 questions per sheet. We developed a system that generates and trains on diverse OMR templates, allowing it to process millions of variations. Using dynamic clustering, the system thoroughly analyzes the filled sheets, detecting answer patterns and verifying whether a response is valid. The dynamic OMR reader leverages AI to understand and adapt to different human patterns when filling OMR sheets for exams. It processes a wide range of sheet formats by learning from customized templates, ensuring seamless handling of various exam types. By integrating machine learning, the system can now recognize true or false answers with precision, adapting to the unique ways in which individuals fill their sheets.
Business type:
Technologies:
Industry:
1
Inaccuracy of the Old System
The previous system owned by the client was inaccurate.
2
Time-Consuming
The existing process took 2 seconds per sheet, making it slow.
3
Lack of Al Components
The client's system could not recognize handwritten digits, leading to verification issues.
4
Variability in Human Patterns
Human variability in filling OMR sheets leads to inconsistencies in marking and alignment, and the dynamic clustering technique helps adapt to these patterns.
Users can easily upload and process various exam formats, from small tests with 50 questions to larger ones with 300 questions or more.
The system can process multiple OMR sheets, at one go, reducing the time needed to evaluate large volumes of answer sheets.
The OMR reader accurately verifies if bubbles are properly filled, ensuring correct marking and reducing the chance of errors in result declarations.
By leveraging AI to detect patterns, the system ensures that all filled sheets are thoroughly analyzed, providing accurate, error-free results to colleges.
1
Template Builder
2
Bulk Sheet Processing
3
Handwriting Recognition
4
Customizable Templates
Developed Convolutional Neural Network (CNN) models for high-precision Optical Character Recognition (OCR) to accurately extract handwritten roll numbers.
Utilized clustering algorithms to verify if bubbles were properly filled or empty, ensuring accuracy in the marking process.
Our product can process 2-3 OMR sheets per second.
Implemented computer vision techniques to automatically detect all bubbles on OMR sheets, eliminating manual effort.
Added a barcode detection feature for additional verification and data integrity.
The new dynamic OMR reader system significantly enhanced the speed and accuracy of processing exam sheets. It reduced time complexity by processing more sheets in less time and successfully addressed the issues of handwritten digit recognition and bubble verification. As a result, the client now enjoys a faster, more reliable system that handles vast exam formats with ease.
Empowering Educators with Fast, Flexible OMR Analysis
