A simple Sudoku solver made in python 3.6
Note: Try to keep the source, i.e the paper flat. Else, the OCR might not work.
For simplicity, you can download the executable from the Google Drive link in drive.txt
NOTE: This only works on Windows. For that, you need to have Tesseract Installed. The application doesn't work without Tesseract Install Tesseract at
C:/Program Files (x86)/Tesseract-OCR
The source code is in master_sudoku.py
The following dependencies are required:
numpy
opencv
pygame
pytesseract
Alternatively, you can run:
pip install -r requirements.txt
Also, you should download the Tesseract OCR. Here is the link for it: Tesseract (Source Forge)
If you are using the souce code, you should update the path of the Tesseract OCR in the code. I have commented the instructions for further details.
I have used Tesseract OCR for image recognition. If you want, you can train a custom Tensorflow Model and import it for character recognition. The predict_2.py file is one such model. I have not used it since the recognition speed is low. But it has a higher accuracy.
The model requires the model2.ckpt file to work.
For generating the model2.ckpt file, run the training script create_model_2.py by using the command
python create_model_2.py
But first, take the following files:
t10k-images-idx3-ubyte.gz
t10k-labels-idx1-ubyte.gz
train-images-idx3-ubyte.gz
train-labels-idx1-ubyte.gz
and put them into a folder called MNIST_data and then run the script. Else, you can change the path of the files in the create_model_2.py
This project was developed by Rochana HM.