Non-binary output possibility of one-class SVM classifier

I have a training set consisting of 10 20x20 grayscale images. I trained a one-class SVM by

SVMStruct = fitcsvm(trainData,class,'KernelFunction','rbf','Standardize',true,'Nu',1);

Now it perfectly works for classification by: [label,score] = predict(SVMStruct,inputImg)

where I get the score of -1 for each member out of class and 0 for training set and very similar images. However I need to see outputs as percentages of closeness or any results of real numbers instead of binary. Labels are not my concern. I investigated predict.m; but couldn't find any rounding operator.

Is there any way to extract the output values before final and binary classification without changing the parameters I mentioned above??

Thank you