Some question about Shadow Prior Map Images

[Jingwei-Liao]

Have you ever retrained the SVM on your data or you only train the RGBPCNN,we don't understand this in your paper

[Jingwei-Liao]

how can we train the SVM on our own data？ thanks very much