I am building a Visual Recognition model on Watson Studio and time by time I retrain the model by adding more photographs. However recently it seems that the current classes kee
