One approach to increasing the accuracy in a model regardless of the amount of data you have is to create artificial examples based on existing data. This is called data augmentation. Data augmentation can also be used at test time to improve prediction accuracy.