https://github.com/joyrex2001/sudosolv
Sudosolv will take a picture of a sudoku and will solve the sudoku accordingly.
https://github.com/joyrex2001/sudosolv
machinelearning sudoku
Last synced: 4 months ago
JSON representation
Sudosolv will take a picture of a sudoku and will solve the sudoku accordingly.
- Host: GitHub
- URL: https://github.com/joyrex2001/sudosolv
- Owner: joyrex2001
- License: mit
- Created: 2022-05-06T19:59:09.000Z (about 3 years ago)
- Default Branch: master
- Last Pushed: 2022-05-31T11:07:59.000Z (about 3 years ago)
- Last Synced: 2025-01-23T00:28:53.173Z (6 months ago)
- Topics: machinelearning, sudoku
- Language: Go
- Homepage:
- Size: 88.9 KB
- Stars: 1
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# sudosolv
Sudosolv will take a picture of a sudoku and will crop the sudoku from the image and decode this into plaintext.
To use this, you will need to train a number recognition classifier first. There are two datasets possible, either the mnist dataset, or a generated dataset based on available fonts.
Prepare the fonts dataset:
```shell
cd dataset/fonts
sh download.sh
sh create.sh
```
Prepare the mnist dataset:
```shell
cd dataset/mnist
sh download.sh
```
Train the network:
```shell
make train
```
Test with some sudoku image:
```shell
go run main.go decode -w trained.bin -f my_sudoku_image.jpg
go run main.go decode -w trained.bin -f my_sudoku_image.jpg --display
```
Or run the webserver and play around on http://localhost:8080:
```shell
go run main.go server -w trained.bin
```