A drawing game where the player has to draw the specified classes as fast as possible. The computer can guess classes using a Convolutional Neural Network(CNN).

• Run DoodleGuesser.py
• Attempt to draw the doodles as quickly as possible.

• Guessing is done with a Convolutional Neural Network with 93387 parameters
• Model can classify images to 1 of 11 classes with a > 95% accuracy
• Easy to add more classes and recompile model

• Python (Built with Python 3.11.14)
• Tensorflow (Built with TF version 2.9.0)
• Numpy (Built NP version with 1.25.0)
• Pygame (Built with pygame 2.5.0)

1. Install Python 3.11.14
2. Clone the repo

   git clone https://github.com/Mihir-A/Doodle-Guesser.git
   cd Doodle-Guesser

3. Create and activate a virtual environment

   python3 -m venv .venv
   source .venv/bin/activate

4. Install necessary libs

   python3 -m pip install -r requirements.txt

5. Download more class data from Google Quickdraw if desired
6. Use modelCreator.ipynb to create a new model. Required if new classes have been added.
7. Run doodleGuesser.py

   python3 doodleGuesser.py

Distributed under the MIT License. See LICENSE.md for more information.

mihiranan@gmail.com

Project Link: https://github.com/Mihir-A/Doodle-Guesser

• Google Dataset

(back to top)