Question on image conditioning for sparse-view reconstruction

[jychoi118]

Thank you for sharing the code. The work is very interesting.

Have you tried training image-conditioned diffusion model for sparse-view reconstruction?

As I understand the paper, SSDNeRF uses unconditional diffusion model + guidance, instead of a conditional diffusion model. I found concatenation conditioning in the code, but it seems to be unused.

[Lakonik]

Thank you for your interest in our work! Our preliminary results show that vanilla concat conditioning + guidance works better than guidance alone, and can generalize to multiple categories. But this is still at a very early stage so we have not yet included the config for image-conditioned model.

[jychoi118]

Thank you for your response!