A question about making psuedo-gts.

From your paper I understood that you used SMPLify when making psuedo gts.

More specifically, it says that you use 2d gt when using SMPLify on MPII, MSCOCO datasets.

My question is did you apply the SMPLify for every single images or on a batch scale like you do for training?

Thanks a bunch.