Training python train_emotion_classifier.py Testing python run_classifier.py test.jpg Things to try: Look at the data! How is it distributed? How well would a random model do? Can the model learn on a tiny subset of the data? Better models Better architecture Data cleanup Normalize Data generation https://keras.io/preprocessing/image/ Reduce learning rate on plateau https://keras.io/callbacks/#reducelronplateau Find more training data online? Anything else?