Question about DL3DV resolution used in Lyra2 training

[sealzhao]

Hi, thanks for the great work!

I'm trying to better understand/reproduce the training setup.

I noticed that Lyra2 uses the DL3DV dataset during training. Could you clarify which resolution/version of DL3DV was used?

For example:
original resolution?
resized images (e.g. 256 / 512 / 1024)?
any specific cropping strategy?

Thanks!