Main class for SuGaR models. Because SuGaR optimization starts with first optimizing a vanilla Gaussian Splatting model for 7k iterations, we

Class to wrap original Gaussian Splatting models and facilitates both usage and integration with PyTorch3D.

Class to wrap Gaussian Splatting camera parameters and facilitates both usage and integration with PyTorch3D.

Wrapper of the densification functions used for Gaussian Splatting and SuGaR optimization. Largely inspired by the original implementation of the

Wrapper of the Adam optimizer used for SuGaR optimization. Largely inspired by the original implementation of the 3D Gaussian Splatting paper:

Class to store Gaussian Splatting camera parameters.

r"""Creates a criterion that measures Learned Perceptual Image Patch Similarity (LPIPS). Arguments: net_type (str): the network type

Parameters of the Gaussian Splatting model. Largely inspired by the original implementation of the 3D Gaussian Splatting paper: https://github

Parameters of the Gaussian Splatting optimization. Largely inspired by the original implementation of the 3D Gaussian Splatting paper: https:/

Parameters of the Gaussian Splatting pipeline. Largely inspired by the original implementation of the 3D Gaussian Splatting paper: https://git

sort angle / dist combined

Forward libav declarations.

Ambient occlusion options. */

Helper to handle window creation/destruction.

Barycentric coordinates

Specialization of Subview for basic views. */

Default voxel type, storing binary occupancy. */

Camera converter args. */