Built and compared a convolutional neural network, feed-forward neural network and SVM to predict google's quick draw game out of 31 categories. Best achieved is a score of 82% using CNN.
Built in Keras/Tensorflow, and preprocessed data with OpenCV 3.0.