Solve sudoku puzzle taking an image as input. Each digit can be handwritten. Note: currently the code only supports puzzle images which the largest outliner and the outliner of each 3 * 3 blob are both bold black, and the inliner of each 3 * 3 blob are light black. Check the manual3.bmp for an example.
- Use OpenCV to locate each unit-blob
- Use TensorFlow to build a MNIST digit recognizer and recognize the digit in each unit-blob
- Use dlxsudoku package to solve extracted puzzle
- Anaconda 3
- Python 3.6