Hi, thank you for your work and sharing.
Would you please clarify why we optimize the scene codes? My point is the NeRF decoder can work on the denoised codes from the diffusion model. If we do so, the model only need to learn diffusion prior (pϕ(x)) and NeRF params. As such, why is learning scene codes required?
Thanks once again.
Regards.
Hi, thank you for your work and sharing. Would you please clarify why we optimize the scene codes? My point is the NeRF decoder can work on the denoised codes from the diffusion model. If we do so, the model only need to learn diffusion prior (pϕ(x)) and NeRF params. As such, why is learning scene codes required? Thanks once again. Regards.