I can't replicate paper results

[gnicolasb]

Hi,

First of all, thanks for sharing this amazing work

I'm trying to replicate the Wonder3D++ results with the Dragon2.png example, but the results are very far from the results shown in the paper. I'm not sure what I'm doing wrong, but I think it is related to the geometric initialization stage. This is the result I get:

Thank you!