About how to calculate deception score for our own model of our own style dataset

[xurong1981]

I can evaluate the new models of the artists mentioned in github page, based on evaluation data provided by the authors. But I wonder how to calculate deception score for new models of new styles, for example, I collect a data set on ink painting, and made a new model of ink painting. For this case, how can I calculate deception score for the new model of ink painting ?

Thank you so much for your comments.