“Doodle Dash” is an engaging machine learning (ML) powered web game created by Xenova, which operates completely in a user’s browser, thanks to the use of Transformers.js. Hosted on the Hugging Face Spaces platform, this web game challenges users to draw as many words as they can in a span of 60 seconds while a neural network attempts to guess the drawings in real-time.
The game is inspired by Google’s “Quick, Draw!” game, where users are given a word and a neural network is tasked to guess the drawn representation of the word. However, Doodle Dash provides a twist: the user is given one minute to draw as many items as possible, one prompt at a time. Once the ML model successfully predicts the correct label for a drawing, the canvas is cleared and a new word is provided for the user to draw [1].
The code and assets for Doodle Dash are hosted on Github [2], where they are actively maintained by the Hugging Face staff [3]. The platform allows users to report bugs, suggest improvements, and participate in discussions about Doodle Dash. For instance, users have suggested features like adding a leaderboard and additional time to the clock when an object is correctly recognized. This encourages an open source collaborative approach to the game’s development and enhancement.
Overall, Doodle Dash presents an innovative and interactive way of leveraging machine learning in a gaming context, promoting learning, and encouraging collaboration within the developer community.
