The proposed multi-view cross-domain diffusion scheme enhances 3D generation by ensuring geometrical and visual coherence across different views, leveraging large foundation models.
Using multi-view attentions, our method generates multi-view normal maps and color images while enforcing consistency, addressing the limitations of prior models that produced separate, inconsistent outputs.
Collection
[
|
...
]