Cross-dataset evaluation

[XpracticeYSKM]

Can you share the scripts for cross-dataset evaluation， it seems that there only has base-to-novel evaluation

[lixinustc]

Of course, I will provide it in this repo, thanks.

[XpracticeYSKM]

And I want to figure out the results of Figure 3 is trained on 'all' or 'base' class? I don't find 'SUBSAMPLE_CLASSES' is set to 'new' in code repo.