Optical Character Recognition (OCR) has advanced significantly with deep learning, yet automatically recognizing and parsing printed Sudoku grids from low-quality physical media-such as newspapers and magazines-remains a persistent challenge due to variable typography, perspective distortions, localized shadows, and ink degradation. This paper presents a complete, web-based end-to-end software pipeline designed to localize, digitize, and interactively guide users through solving printed Sudoku puzzles. The proposed system features an robust five-strategy cascade for image thresholding and grid localization, ensuring structural invariance against non-uniform illumination and perspective skewing. Following geometric rectification via homographic perspective transformation, individual cells undergo a customized twelve-candidate binarization workflow reinforced by Contrast Limited Adaptive Histogram Equalization (CLAHE) to effectively preserve thin or faint character boundaries. Character recognition is driven by a deep convolutional neural network architecture, termed DigitCNN, integrating channel-wise Squeeze-and-Excitation block mechanics and symmetric identity skip connections to automatically capture spatial hierarchies of printed digits across diverse typefaces. To narrow the persistent domain gap encountered in document analysis pipelines, the DigitCNN model is optimized using an innovative self-supervised, pipeline-aware training regime spanning 1.62 million stochastically degraded synthetic and augmented text samples processed through the exact binarization pipeline executed during inference. To overcome individual classification errors, the system embeds a post-classification constraint correction layer acting on global puzzle logic, combined with a self-supervised template rescoring module utilizing high-confidence local predictions to re-evaluate ambiguous cells. Experimental evaluations demonstrate that the customized DigitCNN model achieves a peak validation accuracy of 99.04% and a final convergent accuracy of 98.91%, vastly outperforming classic structural frameworks. Integrated into a monolithic Single Page Application (SPA) driven by a Flask-backed runtime architecture, the system incorporates a non-deterministic Minimum Remaining Values (MRV) heuristic-based backtracking solver to parse real-time uniqueness violations dynamically. Rather than acting as an automated solver that bypasses human reasoning, the application operates strictly as an intelligent hint provider-highlighting conflicting nodes, maintaining dual pencil-and-pen entry modes, and exposing logical per-cell constraints directly within the web interface to augment human deductive problem-solving.
Optical Character Recognition (OCR), Convolutional Neural Networks (CNN), Sudoku Grid Detection, Squeeze-and-Excitation (SE), Perspective Correction, Heuristic Backtracking Solver, Human-AI Interaction.
International Journal of Trend in Scientific Research and Development - IJTSRD having
online ISSN 2456-6470. IJTSRD is a leading Open Access, Peer-Reviewed International
Journal which provides rapid publication of your research articles and aims to promote
the theory and practice along with knowledge sharing between researchers, developers,
engineers, students, and practitioners working in and around the world in many areas
like Sciences, Technology, Innovation, Engineering, Agriculture, Management and
many more and it is recommended by all Universities, review articles and short communications
in all subjects. IJTSRD running an International Journal who are proving quality
publication of peer reviewed and refereed international journals from diverse fields
that emphasizes new research, development and their applications. IJTSRD provides
an online access to exchange your research work, technical notes & surveying results
among professionals throughout the world in e-journals. IJTSRD is a fastest growing
and dynamic professional organization. The aim of this organization is to provide
access not only to world class research resources, but through its professionals
aim to bring in a significant transformation in the real of open access journals
and online publishing.