🛠 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.

🕵️ Introduction

Remember the summer vacation nostalgia of sitting down to solve a giant 200 piece puzzle? At some point, all of us enjoyed and solved puzzles. The panic of a missing piece and the joy of completing puzzles was exciting!

This challenge will try to recreate the retro-puzzle solving with an AI twist!

Given a set of jumbled images, can you sort them in the correct order to solve the puzzle and form the complete picture?

The mission, if you choose to accept, is to classify and sort the many jumbled images in the correct order and submit a final solution with the correct order number.

Understand with code! Here is getting started code for you.😄

💾 Dataset

The dataset contains puzzle pieces for an images, inside the folder with name as the image id in puzzles.tar.gz. The height and widhth of original image can be found in the metadata.csv . There are total 2500 such folders, with puzzle pieces of the respective image ids.

📁 Files

Following files are available in the resources section:

• puzzles.tar.gz: A tar.gz file which when extracted, has 2500 folders (where folder names are the puzzle-id), and each of the folders contain the individual puzzle pieces comprising this puzzle as PNG files.

• sample_submission.tar.gz: A tar.gz file with 2500 randomly reconstructed images of each of the puzzles in the test set. The naming convention for each of the files inside this tar is <puzzle_id>.jpg.

• metadata.csv : A file containing the width and height of each of the puzzles in the test set.

🚀 Submission

• Recreate the original images with the puzzle pieces given in each folder for an image.

• Name the image as {image_id}.jpg

• Create a tar.gz file containing all the recreated images.

• For eg       

• Sample_submission.tar.gz can be found in resources section.

Make your first submission here 🚀 !!

🖊 Evaluation Criteria

This challenge uses the SSIM score as the primary evaluation metric and the Mean Squared Error.

For all the puzzles in the dataset, the individual scores are computed by comparing the submitted reconstructed image in reference to the original image. The overall submission score is the mean SSIM and MSE scores across the whole data

The score of only 60% of the test data will be revealed during the competition.

🔗 Links

• 💪 Challenge Page: https://www.aicrowd.com/challenges/jigsaw
• 🗣️ Discussion Forum: https://www.aicrowd.com/challenges/jigsaw/discussion
• 🏆 Leaderboard: https://www.aicrowd.com/challenges/jigsaw/leaderboards

📱 Contact

• Sharada Mohanty