Before anything else, I'd like to thank the random Samaritan on the internet who was kind enough to provide a detailed article on the build-up of the project. It was immensely helpful. Shoutout to McGill School of Computer Science for providing the Sudoku algorithm.

The GitHub repository hosts an Image-based Sudoku Solver project, combining Computer Vision and Machine Learning. Leveraging OpenCV for image processing and TensorFlow for ML integration, the solver captures real-time video frames to recognize and solve Sudoku puzzles. The repository includes code for contour detection, perspective transformation, and grid recognition, showcasing a comprehensive solution. A pre-trained neural network facilitates accurate digit recognition, making the solver adept at handling diverse puzzles. The project is well-documented, offering insights into its implementation and potential extensions. The integration of Tkinter for a user-friendly interface adds to its accessibility and usability.

This program serves as a way to calculate the solution to any 9x9 sudoku puzzle via webcam. It identifies the puzzle through the webcam, processes it uses OpenCV, runs against a neural network to predict the digits, and runs an efficient sudoku solver to determine the answer. It then displays the answer on the same frame if it is solvable.

PuzzlVision is an innovative project that combines computer vision and machine learning to create an image-based Sudoku solver. The system captures real-time video frames, processes Sudoku grids, and employs a pre-trained neural network for accurate digit recognition, ultimately solving Sudoku puzzles.

Tested using Python 3.6 (newer versions may or may not work)

• opencv-python: 4.3.0.36
• numpy: 1.19.1
• tensorflow: 2.2.0
• sklearn: 0.0
• keras: 2.3.1

• Real-Time Sudoku Solver: Capture and process video frames to solve Sudoku puzzles in real-time.
• Machine Learning Integration: Utilize a pre-trained neural network for accurate digit recognition.
• Comprehensive Workflow: Include contour detection, perspective transformation, and grid recognition in the image processing pipeline.
• User-Friendly Interface: Potential integration with Tkinter for an intuitive user interface for puzzle input and solution display.

• OpenCV: For image processing and contour detection.
• TensorFlow: For machine learning model integration.
• Neural Network: A pre-trained model for accurate digit recognition.
• Tkinter: (Optional) For creating a user-friendly interface.

• helpers/: Contains auxiliary scripts and utility functions used across the project.
• current/: Hosts the main source code for the image-based Sudoku solver. Mainly the primary algorithm and process segregations.
• original/: Stores the original, unprocessed images for reference and comparison.
• preprocessing/: Includes image preprocessing scripts to enhance image quality before digit recognition.
• models/: Holds pre-trained machine learning models for accurate digit recognition.

1. Clone the repository:

   git clone https://github.com/your-username/puzzlvision.git
   cd puzzlvision

2. Run the main script:

   python app.py

• Capture Input: Point the camera towards a Sudoku puzzle.
• Real-Time Processing: The system processes video frames and displays the solution.
• User Interface (Optional): If using Tkinter, interact with the user-friendly interface for puzzle input.

This project is licensed under the MIT Licence.

• Credits to the Sudoku Solver Algorithm
• OpenCV Documentation
• TensorFlow Documentation
• McGill School of Computer Science
• ChatGPT