search

Anttwo / SuGaR

[CVPR 2024] Official PyTorch implementation of SuGaR: Surface-Aligned Gaussian Splatting for Efficient 3D Mesh Reconstruction and High-Quality Mesh Rendering
https://anttwo.github.io/sugar/
Other
2.36k stars 182 forks source link

Can the author explain what the "get_field_values" function does? #65

Open wangyuanbiubiubiu opened 11 months ago

wangyuanbiubiubiu commented 11 months ago

def get_field_values(self, x, gaussian_idx=None, closest_gaussians_idx=None, gaussian_strengths=None, gaussian_centers=None, gaussian_inv_scaled_rotation=None, return_sdf=True, density_threshold=1., density_factor=1., return_sdf_grad=False, sdf_grad_max_value=10., opacity_min_clamp=1e-16, return_closest_gaussian_opacities=False, return_beta=False,): I can't understand this function very well, can the author provide some help？

wangyuanbiubiubiu commented 11 months ago

In fact, I have been puzzled by a sentence in the article, why the density function is determined by the Gaussian value, rather than the weight value of T*a

wangyuanbiubiubiu commented 11 months ago

sdf_estimation_loss = (sdf_values - sdf_estimation.abs()).abs() / sdf_sample_std Here, both sdf_values and sdf_estimation actually return the gradients, is that what's expected? One should be an measurement value