r""" class for ssim Args: win_size: (int, optional): the size of gauss kernel win_sigma: (float, optional): sigma of n

Generate a 360 degree spherical path for rendering.

Generate a spiral path for rendering.

Average poses according to the original NeRF code.

Recenter poses according to the original NeRF code.

Recenter poses according to the original NeRF code.

Initialize training.

Construct lookat view matrix.

Computes the exposure map corresponding to the HDR-FLIP error map :param hdrflip: matrix (with HxW layout) containing per-pixel HDR-FLIP errors :

Computes start and stop exposure for HDR-FLIP based on given tone mapper and reference image. Refer to the Visualizing Errors in Rendered High Dyna

[H, W, C], [0, 1]

Computes the FLIP error map between two HDR images, assuming the images are observed at a certain number of pixels per degree of visual angle :p

[B, C, H, W], [0, 1]

[B, C, H, W], [0, 1]

[B, C, H, W], [0, 1]

Args: - x: np.ndarray, [B, 3] Returns: np.ndarray, [B, 1]

Convert a set of rays to NDC coordinates.

Eases in each frequency one by one with a cosine. This is equivalent to taking a Tukey window and sliding it to the right along the frequency spec

Return the -std=c++[11/14/17] compiler flag. The newer version is prefered over c++11 (when it is available).

Assuming `depth` is orthographic, linearize it to a set of normals.

Estimates mean and variance of sin(z), z ~ N(x, var).

Helper to find or download the Eigen C++ library

Frees all allocated memory chunks */

Return the viridis map, described here: https://bids.github.io/colormap/ :return: matrix of size 256x3 where each row contains a color in the vird

* Returns a pointer to a piece of new memory of the given size in bytes * allocated from the pool. */

r""" interface of ms-ssim Args: X (torch.Tensor): a batch of images, (N,C,H,W) Y (torch.Tensor): a batch of images, (N,C,H,W)

Compute overlapping weighted FLIP histogram of two error maps :param flip_error_map_array: matrix array (with HxWx2 layout) containing per-pixel F

Prints information about the metric invoked by FLIP :param pixels_per_degree: float indicating number of pixels per degree of visual angle :param

Prints pooled values of the FLIP error map :param error_map: matrix (with HxW layout) containing per-pixel FLIP values in the [0,1] range :param

Args - pts: np.ndarray, [H, W, D, 4] - cam_mat: np.ndarray, [4, 4] - view_mat: np.ndarray, [4, 4] Returns: - uv: np.ndarray, [H,

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

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

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

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

see: src/base/reconstruction.cc void Reconstruction::ReadPoints3DText(const std::string& path) void Reconstruction::WritePoints3D

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

Redistributes exponentiated HyAB errors to the [0,1] range :param power_deltaE_hyab: float containing the exponentiated HyAb distance :param cmax

jnp.norm output nan gradient if 0

Determine the observer's number of pixels per degree of visual angle based on input arguments :param args: set of command line arguments (see main

Split data into shards for multiple devices along the first dimension.

Piecewise-Constant PDF sampling from sorted bins. Args: key: jnp.ndarray(float32), [2,], random number generator. bins: jnp.ndarray(float32

Args: - exp: float, 0 for cosine, 1 for uniform distribution

r""" interface of ssim Args: X (torch.Tensor): a batch of images, (N,C,H,W) Y (torch.Tensor): a batch of images, (N,C,H,W)

Args: - rvec: np.ndarray, [3, 1] - tvec: np.ndarray, [3, 1] Returns: - mat: np.ndarray, [4, 4]

Args: - mat: np.ndarray, [4, 4] Returns: - ret: np.ndarray, [4, 4]

Collect the sharded tensor to the shape before sharding.