Sparse-multiview-input supporting comes soon!

[HpWang-whu]

TODO: Allow users to input more views and no poses are needed. This feature will be updated at about December.
An example result of VistaDream with two unposed images as input are shown below.

Inputs:

RGB renderings from the scene reconstructed with VistaDream:

refine.video_rgb.mp4

[yqdch]

Cool, if this video was rendered using optimized 3DGS