Abstract: Synthesizing novel views from sparse views has achieved impressive advances with radiance fields, yet prevailing methods suffer from high consumption or insufficient refinement capability.
In this project, we are implementing a visual computing pipeline to reconstruct a 3D scene from a multi-view image sequence using 3D Gaussian Splatting. The system will transform 2D image frames into ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results