Return the masked tensor along with optional loss/metrics.

Decodes a bitstream file to a YUV file. Args: input_bitstream: Path to the input bitstream file. output_yuv:

Generates standard file paths for an attribute component.

Entry for running video anchor compression.

Read and unpack the next bytes from a binary file. :param fid: :param num_bytes: Sum of combination of {2, 4, 8}, e.g. 2, 6, 16, 30, etc.

Read and unpack the next bytes from a binary file. :param fid: :param num_bytes: Sum of combination of {2, 4, 8}, e.g. 2, 6, 16, 30, etc.

Wrapper to run a function in a distributed environment. The function `fn` should have the following signature: ```python def fn(local_ra

Rasterize a set of 3D Gaussians (N) to a batch of image planes (C). This function provides a handful features for 3D Gaussian rasterization, whic

Initialize and return the running state for this strategy.

Save a tensor to a YUV file in planar format. Args: video (torch.Tensor): video in the shape of [T, H, W, 3], assumed to be uint8 an

Sort splats with Parallel Linear Assignment Sorting from the paper `Compact 3D Scene Representation via Self-Organizing Gaussian Grids <https://ar

Callback function to be executed after the `loss.backward()` call.

Load a raw YUV file into a PyTorch tensor. Reads planar YUV data and reconstructs it into a tensor of shape [T, H, W, 3] with dtype uint

Update the parameters and the state in the optimizers with defined functions. Args: param_fn: A function that takes the name of the param

Decompresses data from a compressed directory. 1. Decodes each bitstream back to a YUV file. 2. Loads data from YUV files us

Projects Gaussians to 2D. This function fuse the process of computing covariances (:func:`quat_scale_to_covar_preci()`), transforming to came

Converts quaternions and scales to covariance and precision matrices. Args: quats: Quaternions (No need to be normalized). [N, 4]

Callback function to be executed before the `loss.backward()` call.

Split and exchange tensors between all ranks in a many-to-many fashion. Args: world_size: The total number of ranks. tensor_list:

Decode video using OpenCV and handle color channels based on parameter name.

Decode video using OpenCV and handle color channels based on parameter name.

Encodes intersection ids to offsets. Args: isect_ids: Intersection ids. [n_isects] n_cameras: Number of cameras. tile_wid

Maps projected Gaussians to intersecting tiles. Args: means2d: Projected Gaussian means. [C, N, 2] if packed is False, [nnz, 2] if packed

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

Transform cameras using an SE(3) matrix. Args: matrix: 4x4 SE(3) matrix camtoworlds: Nx4x4 array of camera-to-world matrices

Transform points using an SE(3) matrix. Args: matrix: 4x4 SE(3) matrix points: Nx3 array of points Returns: Nx3 arra

Quantizes a tensor to a specified bit depth after reshaping and normalization.

Sanity check for the parameters and optimizers.

Convert byte size to readable format (KB, MB, GB, etc.)

Prepare Gaussians for rasterization This function prepares ray-splat intersection matrices, computes per splat bounding box and 2D means in i

Observe the current splats before running compression.

Inverse transform the splats parameters to original domain Args: splats_list (List[Dict]): splats parameters of all frame

Transform the splats parameters to new domain e.g. for SH coefficients, transform them from RGB domain to YCbCr domain for q

see: src/base/reconstruction.cc void Reconstruction::ReadImagesBinary(const std::string& path) void Reconstruction::WriteImagesBi

Taken from https://github.com/colmap/colmap/blob/dev/scripts/python/read_write_model.py

see: src/base/reconstruction.cc void Reconstruction::WriteCamerasBinary(const std::string& path) void Reconstruction::ReadCameras

Taken from https://github.com/colmap/colmap/blob/dev/scripts/python/read_write_model.py

Computes spherical harmonics. Args: degrees_to_use: The degree to be used. dirs: Directions. [..., 3] coeffs: Coefficient

Returns total variation on multi-dimensional tensors. Args: x (torch.Tensor): The input tensor with shape $(B, C, ...)$, where $B$ is the

Construct lookat view matrix.

Dequantizes a video tensor using metadata.

Exchange 32-bit integers between all ranks in a many-to-many fashion. .. note:: This function is not differentiable to the input tensors.

Sanity check for the parameters and optimizers. Check if: * `params` and `optimizers` have the same keys. * Each opti

Entry for evaluation.

Creates a smooth spline path between input keyframe camera poses. Spline is calculated with poses in format (position, lookat-point, up-point).

see: src/base/reconstruction.cc void Reconstruction::ReadPoints3DBinary(const std::string& path) void Reconstruction::WritePoints

Callback function to be executed after the `loss.backward()` call. Args: lr (float): Learning rate for "means" attribute of the G

Recursively get relative paths of files in a directory.