Regarding the rendering method of the expected depth in the rasterizer #7
Description
I noticed that the rasterizer you use is different from the accumulated depth in Deformable 3D Gaussian Splatting, or the inverse depth in the original 3DGS. Instead, you compute the expected depth by performing an opacity-weighted normalization.
May I ask what the motivation is behind using the expected depth?
Is it intended to better approximate the true depth of the scene?