Sudoku Savant

Example Sudoku Puzzle
Finished Puzzle Solved with Sudoku Savant

Inspiration

Sudoku Savant was inspired by the desire to build a simple and efficient method for solving Sudoku problems using image recognition and machine learning. We intended to eliminate the necessity for manual solutions and provide Sudoku fans with a seamless experience.

What it does

Sudoku Savant is a Python application that solves Sudoku puzzles given an image using a TensorFlow model. It can evaluate the puzzle grid from the image and deliver the solution in seconds by leveraging advanced image recognition techniques and powerful machine learning algorithms.

How we built it

To build Sudoku Savant, we utilized Python as the primary programming language. We leveraged the TensorFlow framework to train a model capable of recognizing and understanding Sudoku puzzle grids from images. The model was trained on a large dataset of Sudoku puzzles to ensure accurate and reliable results. We also implemented image preprocessing techniques and integrated the model with the Sudoku solving logic.

Challenges we ran into

We experienced various problems when developing Sudoku Savant. To ensure best performance, training the TensorFlow model required careful data selection and preprocessing. We also had trouble correctly extracting the puzzle grid from the input image since differences in lighting, orientation, and quality could impair identification accuracy. In addition, optimizing the method for efficiency and accuracy was difficult when integrating the model with the Sudoku solving logic.

Accomplishments that we're proud of

We take pride in the accuracy and quickness with which Sudoku Savant offers solutions, allowing users to solve problems with ease. Sudoku Savant also accounts for skewed images, to ensure that any sudoku puzzle can be solved.

What we learned

We received great insights into picture recognition techniques, machine learning algorithms, and the integration of these technologies into practical applications while developing Sudoku Savant. We discovered the difficulties of training a model with little data and the significance of preprocessing strategies in improving recognition accuracy. We also improved our abilities in optimizing algorithms for efficiency and performance, and familiarized ourselves with TensorFlow.

What's next for Sudoku Savant

In the future, we see various intriguing possibilities for Sudoku Savant. We intend to improve the program by adding features such as a real-time camera capture mode, which will allow users to solve puzzles directly through their device's camera. We also plan to include advanced solving tactics and hints to help people improve their Sudoku skills. Furthermore, we will continue to refine the model and algorithm in order to increase recognition accuracy and expand compatibility with various puzzle varieties. Last but not least, the capabilities can be implemented into a web app for greater accessibility.