Convenient function to render a point cloud with pointersect. Args: input_point_cloud: the filename of a ply file contai

reset the point representing inf at n=0

set all points as valid

Given a dict or a list containing tensor, reshape dimension `start` to `end` to the new shape

Reshape each attributes.

Sample spherical gaussian distributions. Args: sample_shape: (*sample_shape). Can be an empty list. Returns:

Sample new points from the current point cloud. Args: num_points: number of points to sample method:

Sample from `arr` a patch centered at `center` with a different pixel pitch and number of pixels. Args: arr: (*b, c, h_i

Sample a set of points (with color) on the mesh. Args: mesh: a triangle mesh representing the surface. We as

Samples random uv coordinates ([0, w), [0, h)) to create random patches. Args: b_shape: (*b,) determines the number of p

Generate rays to form patches on the corresponding images. Args: num_patches_per_q: number of patches fr

Used in PCD version of inverse rendering. Args: cam_poses: intrinsics: sample_center: region_width_px:

Save the point cloud as a ply file and a npz.

Save as the format used by LLFF. LLFF processes the outputs of COLMAP, and we will only output a subset of it. See: https:/

Save as the format used by NPBG++ (https://github.com/rakhimovv/npbgpp). We use the DTU dataset's format: - image: 000000.pn

Save as the format used in the RTMV dataset https://www.cs.umd.edu/~mmeshry/projects/rtmv/. Each scene in the dataset contains

Backup the code in a src directory to a dst directory. The function recursively copies the code (determined by their file extensions) in the fold

Save image tensor as numpy and optionally gif and png Args: imgs: (b, h, w, 3) save_gif: whether to s

Args: dataset_lengths: a list containing the length of each data global_idxs: global

Use the hit_map and the camera in the rgbd_image to invalid points. It marks the valid_mask to be False for point that should be carv

Converts an image from sRGB to Linear colorspace. We treat each pixel independently (to color channels as well). Args: x:

Convenient function to call skimage's ssim. Not differentiable. Args: rgb: (h, w, 3), in the range of [0, 1] gts: (h, w, 3)

Apply `func(tensor)` to any tensors in the list or dict. Args: func: function that takes a tensor and outputs a tensor

Test cpu-only version.

Test cuda version.

Test cuda distributed version.

single dataset

two datasets of different lengths, with ratio = 1

two datasets of different lengths, with ratio = 0.3 and 0.5

two datasets of different lengths, with ratio = 0.3 and 0.5, max_sample smaller than len(dataset) * ratio