Join us for the closing webinar where top participants will discuss their solutions!
🛠 Contribute: Found a typo? Or any other change in the description that you would like to see? Please consider sending us a pull request in the public repo of the challenge here.
How soon can you solve a Sudoku? Do you think you can teach your neural net to solve it? 😉 What if it had to solve it the same way that you do? Through vision?
In this challenge, solve Sudokus with a twist! You won’t just solve a Sudoku but learn important fundamentals of Computer Vision too!
This puzzle teaches a lot of things, and addresses SDG 4 - Education at all ages!
Start solving! Here is a getting started code for you.😄
Understand with code! Here is
getting started code for you.
The dataset consists of 10001 images. 5000 images in the training set and 5001 in the test set.
Each image consists of an unsolved sudoku puzzle. Each puzzle is of a
different color, is
rotated by an
angle, has varying
line widths, and the digits all look different. Each image has the dimension
(300, 300) but due to the rotation, the puzzle sizes may vary.
labels.csv contains columns
solution. The field
id corresponds to the name of the file inside the
images directory whereas the field
solution is a string, 81 characters long which is the fully solved puzzle in
row major form.
For example, this row
0385,971683245562479831483152679854267193219538764637914582148795326396821457725346918 says that the puzzle in
images/0385.png has the solution
The following files can be found in the
5000) This file contains a folder
datawith all the images of the sudoku puzzles and a
label.csvwith labels i.e. solutions ( in string format ) for the corresponding puzzle.
5001) This file contains a folder
datacontaining all the images of the sudoku puzzle whose solutions needs to be found.
sample_submission.csv: Sample submission format for the challenge.
- Prepare a CSV containing headers
[id, solution]. This file should contain solutions for all the puzzles given in the test dataset. The description for
solutioncan be found above.
- Name of the above file should be submission.csv.
- Sample submission format can be found in resources section.
Make your first submission here!! 🚀
🖊 Evaluation Criteria
During evaluation accuracy will be used to test the efficiency of the model. A solution for a puzzle will be considered correct only if all 81 values match, i.e., there is no partial marking.
- 💪 Challenge Page: https://www.aicrowd.com/challenges/SUDOKU
- 🗣️ Discussion Forum: https://www.aicrowd.com/challenges/SUDOKU/discussion
- 🏆 Leaderboard: https://www.aicrowd.com/challenges/SUDOKU/leaderboards
- The sudoku puzzles were generated using this.