This is a clone project of the popular word guessing Wordle.

👾 Play the demo here 🕹️

How to play 🕹️

Each day a random image is selected from the ImageNet-Sketch dataset and the goal is to guess one of the labels associated with the image in 6 tries.

Game simulation

• Visionle chooses the following image 🐈 and its associated challenge word will be tabbycat.
• Users will be presented with 8 empty cells and 6 tries to figure out the challenge word tabbycat
• Users can input any combination of letters equal or shorter than 8 letters. This is different from the typical game setup of Wordle and we refer to this play mode by Looney Mode 🤪
• The Hard Mode 🏋️ restricts the users to input only variation of words from the whole set of labels found in the ImageNet-Sketch dataset.
• Example of game run with tabbycat challenge is presented below:
  • Entry screen 💻
  • Guessing with trial and error 🤔. Notice that you can use shorter words and non valid words (i.e. any letter combination)
  • Success 🎉
  • Stats screen 📊 after the end of trials
  • Sharing results by clicking on Share

Extension to support ML model prediction

The current version of Visionle can be easily extended/modified to use the prediction of a trained Machine learning model such as EfficentNet or Vit on the ImageNet dataset. In fact, the code/logic of this setup is already implemented - what is needed is some slight modification of word.ts file and dumping the predictions of the model in a json file. I will leave this as an exercise for the reader 😅 (until I get the time to push that in version 0.2).

Credits

This repo is build on the code from react-wordle (awesome open source version of Wordle).