Whether to add a mask to the sky in the mipnerf360 dataset?

[OUC-voyage]

For the mipnerf360 data set, did the author add a mask to the sky of the scene? The depth of the sky is often estimated inaccurately. The sky is easy to adopt a progressive propagation strategy, which easily generates a large number of Gaussians, which has an impact on the final depth and normals.

[kcheng1021]

Yes, the sky needs mask. The sky is close to infinitely distant, with almost no parallax in the image, and propagation cannot estimate the correct depth.