- Description
- Design
- Lib used
- Notes
Sudoku solver is a little project to help me resolve my sudoku. I was bored to not having the solution of my sudoku. So i end up doing this to prepare the solution before doing my sudoku.
It help me to re-do a little project with Python, Web and AWS technologies.
So in the end, what you have is:
- A web with Bootstrap
- A login management with AWS Cognito
- An upload page where you can push your sudoku picture and after a processing done by AWS Lambda - Restful API call the image you've uploaded with the right numbers filled.
- The back-end use AWS lambda, AWS Api gateway and Python
Overview of the solution:
- Bootstrap 4.4
- Jquery 3.4.1
- Popper 1.16
- AWS Cognito SDK
- API Gateway
- AWS Lambda & AWS Lambda Layer
- AWS S3 static hosting
- AWS IAM, AWS Cognito
- AWS S3 for layers
- For the image recognition, i follow the help of this github: https://github.com/abidrahmank/OpenCV2-Python
Lib used:
- Numpy
- OpenCV
- Pillow
- Pytesseract
Help for the parts i've got problems:
When i first re-use API Gateway and Lambda for this project, i've got the CORS bad return. To solve this problem do not forget two important part on the "Server" side AWS:
- On your api gateway ressource, use the option to "enable CORS", and after do not forget to "redeploy"
- On your lambda code, do not forget to return in your response the CORS flag. This is the part where is was really stuck:
This is a non-working example:
return { 'statusCode': 200, 'body': json.dumps("TEST") }
This is a working example:
return { 'statusCode': 200, 'headers': { 'Access-Control-Allow-Origin': '*' }, 'body': json.dumps("TEST") }
You can see in the image processing multiple try to be sure of the number detected. At first i was using only one Knearest call but that didn't work for numbers like: 7 <-> 1 and 8 <-> 3
SO i order to improve that, i use multiple blur type and multiple Knearest call and get the max result
- Set up a AWS CloudFormation in order to automate deployment
- Set up a CI with 5 image to test new code
- Set up a CD with AWS to deploy code in the lambda after a new commit or merge in the master branch with github
- Add "Download result" option
- Show side by side orignal and result images
- Clean sudoku python code
- Clean X & Y processing. Too much complicated to retrieve pos X and pos Y from case or square
- Clean list usage to get values