Quick Draw is a well known, AI drawing guessing game that challenges players to draw a doodle, and then, artificial intelligence guesses, what the drawings represent. This tutorial explains how to make it using machine learning and convolutional neural network!
Let’s get started with exploring the finished game which can recognize 3 different doodles: car, snowman, and fish. The main screen is divided into three sections:
- The left section shows the graphs, of accuracy and loss during training, so we can monitor, how the model improves over time.
- The middle section shows the predictions, made on testing samples. Here, the correct predictions are shown in green rectangles, and the incorrect predictions are red.
- The right section is a drawing area, where you can draw your masterpiece with the mouse. Below your drawing, the neural network output is displayed, hopefully showing a correct prediction.
To train the model, press the Train More button.
To test the model using testing samples, press the Next Test button.
Once the model is trained, draw a doodle with your mouse and check the prediction.
Video
To find out how I made this game, here is the video in which I explained the entire project:
Source Code
The complete source code is available at Github:
https://github.com/ssusnic/Machine-Learning-Doodle-Recognition