Text this: Joint representation learning of appearance and motion for abnormal event detection.