Pytorch implementation of `gsplat.cuda._wrapper.rasterize_to_pixels()`. This function rasterizes 2D Gaussians to pixels in a Pytorch-friendly way

Gather an 32-bit integer from all ranks. .. note:: This implementation is faster than using `torch.distributed.all_gather_object`. .

Warp `img` to match the colors in `ref_img` using iterative color matching. This function performs color correction by warping the colors of

Inplace duplicate the Gaussian with the given mask. Args: params: A dictionary of parameters. optimizers: A dictionary of optimiz

Entry for evaluation.

Entry for evaluation.

Entry for evaluation.

Entry for evaluation.

Entry for evaluation.

Generate an elliptical render path based on the given poses.

Calculates a forward facing spiral path for rendering.

cx and cy range is -0.5 to 0.5 we use cx cy range -0.5 * 0.5

Load the test data.

Normalization helper function.

Rasterizes Gaussians to pixels. Args: means2d: Projected Gaussian means. [C, N, 2] if packed is False, [nnz, 2] if packed is True.

Entry for trajectory rendering.

Entry for trajectory rendering.

Inplace reset the opacities to the given post-sigmoid value. Args: params: A dictionary of parameters. optimizers: A dictionary o

reference: nerf-factory Get a similarity transform to normalize dataset from c2w (OpenCV convention) cameras :param c2w: (N, 4) :

Inplace split the Gaussian with the given mask. Args: params: A dictionary of parameters. optimizers: A dictionary of optimizers,

Hook for stepping any optimizers the mask may own.

Convert to dictionary for YAML serialization.

Decodes a video file (e.g., HEVC encoded) to a raw YUV file using FFmpeg. Args: video_file (str): Path to the input video file (.mp4

Encodes a raw YUV file to HEVC (libx265) using FFmpeg. Args: yuv_file_path (str): Path to the input raw YUV file. output_vid

Evaluate spherical harmonics bases at unit direction for high orders using approach described by Efficient Spherical Harmonic Evaluation,

PyTorch implementation of fisheye projection for 3D Gaussians. Args: means: Gaussian means in camera coordinate system. [C, N, 3].

PyTorch implementation of `gsplat.cuda._wrapper.fully_fused_projection()` .. note:: This is a minimal implementation of fully fused vers

Sample from a distribution using torch.multinomial or numpy.random.choice. This function adaptively chooses between `torch.multinomial` and `nump

PyTorch implementation of orthographic projection for 3D Gaussians. Args: means: Gaussian means in camera coordinate system. [C, N, 3].

PyTorch implementation of perspective projection for 3D Gaussians. Args: means: Gaussian means in camera coordinate system. [C, N, 3].

Remove and return the value for the first matching flag in sys.argv.

PyTorch implementation of `gsplat.cuda._wrapper.quat_scale_to_covar_preci()`.

Convert quaternion to rotation matrix.

Executes a subprocess command and handles logging and errors. Args: cmd: The command to execute as a list of strings.

PyTorch implementation of world to camera transformation on Gaussians. Args: means: Gaussian means in world coordinate system. [C, N, 3].

Gather a list of tensors from all ranks. .. note:: This function expects the tensors in the `tensor_list` to have the same shape

Recursively apply ``updates`` onto ``target``.

Chroma downsampling for YUV video from 4:4:4 to 4:2:0 format conceptually, while maintaining the original tensor shape. The U and V value

Applies color affine transformations. Args: affine_mats (torch.Tensor): Affine transformation matrices. Supported shape: $(..., 3, 4)$.

Compute new Gaussians from a set of old Gaussians. This function interprets the Gaussians as samples from a likelihood distribution. It uses

Convert depth maps to surface normals Args: depths: Depth maps [..., H, W, 1] camtoworlds: Camera-to-world transformation matrice

Encodes a YUV file to a bitstream using the specified configuration. Args: input_yuv: Path to the input YUV file.

Entry for evaluation.

Entry for evaluation.

Calculate nearest point to all focal axes in poses.