Thank you for this great work. After trying it out, I found that
- the normal mapping of $\textcolor{red}{\text{moge-3-vitl is not as good as that of moge-2-vitl-normal}}$.
- Is this because $\textcolor{green}{\text{moge-2-vitl-normal is specifically trained on the normal mapping?}}$
If so, do you have plans to open up the weights of moge-3-vitl-normal?
Thank you for this great work. After trying it out, I found that
If so, do you have plans to open up the weights of moge-3-vitl-normal?