Feedback for my Style Transfer Guessing Game

Code Feedback
1

Hey everyone! I recently built a web game where you guess the original content and style images that produced a given stylized image. Inspired by Google’s Arts & Culture games and the concept of style transfer, I wanted to create something interactive that makes learning about style transfer fun.

How it Works:

  • A stylized image is shown on the left.
  • Four images (two correct, two incorrect) appear on the right.
  • You pick the content and style images that generated the stylized image.
  • The game gives immediate feedback with animations, scoring, and visual effects.
  • Try to guess in time and earn points!

Tech Stack & Development:

  • Style Transfer Model: Utilized a PyTorch-based neural style transfer model to generate stylized images.
  • Game Data Storage: Stored generated images & metadata in Cloudinary + JSON on GetPantry (150 rounds of data)
  • Front-end: Built with ReactJS & CSS, with interactive feedback, scoring, and animations.
  • Deployed to Vercel.

There’s an introductory video that I recommend you view when you open the game link, and a help menu is available in the top-right . Hope it helps!

Would love to hear feedback and comments. Thank you.

Play the Game (optimized for laptop)
GitHub Repo

1 Like
3

Hi. It’s neat and it works well even on my slow internet connection. It tends to lose its appeal after a certain amount of rounds. It would be nice if the challenges started to get more difficult as the rounds progress. First thing that comes to mind is adding more pictures to choose from. 6, 8, 10 I don’t know :slight_smile: All in all very nice. I liked it :slight_smile:

1 Like