Feature importance in Forest Models

[elopezfune]

To train any forest model with categorical variables it is needed first to convert them to dummy. After training, to get the feature importance, how does one get only one importance for only one variable, instead of for each class-factor?