How to use it for sparse view reconstruction

[7yzx]

Thanks for your great work , Can it use for sparse view input(3-5views) and poses, then render for other views

[flow0314]

Hi, thanks for your interest in our work! Our method can certainly be used for sparse-view reconstruction (3–5 views).
Since our work is designed to enhance 3DGS quality, you would first reconstruct a base 3DGS from your sparse views and poses. Depending on your goals, you can take two paths:
For better NVS results: Simply render the other views and use GSFixer to restore them.
For better 3D Reconstruction: You can distill those restored results back into the original 3DGS.
Please refer to our README for more detailed implementation steps!