This is for udacity capstone project. A traffic light detector using squeezeNet.
The model (model.h5) is generated using model.save(filepath). It saves the whole model (architecture + weights + optimizer state). The saved model can be loaded by keras.models.load_model(filepath).