kradolfer/quickdraw-image-recognition

Comparing different machine learning algorithms to classify drawings made in the game 'Quick, draw!'.

quickdraw-image-recognition

Comparing different machine learning algorithms in Python to classify drawings made in the game 'Quick, draw!'.

Results:

Classification accuracy for 2 classes (15'000 training examples):

Convolutional Neural Network: 96.4%

K-Nearest Neighbors: 93.6%

Multi-Layer Perceptron: 93.3%

Random Forest: 92.7%

Classification accuracy for 5 classes (12'500 training examples):

Convolutional Neural Network: 90.3%

K-Nearest Neighbors: 82.4%

Multi-Layer Perceptron: 81.2%

Random Forest: 79.1%