Table of Contents
Shortcuts

nerfacc.traverse_grids

nerfacc.traverse_grids(rays_o, rays_d, binaries, aabbs, near_planes=None, far_planes=None, step_size=0.001, cone_angle=0.0, traverse_steps_limit=None, over_allocate=False, rays_mask=None, t_sorted=None, t_indices=None, hits=None)

Ray Traversal within Multiple Grids.

Note

This function is not differentiable to any inputs.

Parameters:

  • rays_o (Tensor) – (n_rays, 3) Ray origins.

  • rays_d (Tensor) – (n_rays, 3) Normalized ray directions.

  • binary_grids – (m, resx, resy, resz) Multiple binary grids with the same resolution.

  • aabbs (Tensor) – (m, 6) Axis-aligned bounding boxes {xmin, ymin, zmin, xmax, ymax, zmax}.

  • near_planes (Optional[Tensor]) – Optional. (n_rays,) Near planes for the traversal to start. Default to 0.

  • far_planes (Optional[Tensor]) – Optional. (n_rays,) Far planes for the traversal to end. Default to infinity.

  • step_size (Optional[float]) – Optional. Step size for ray traversal. Default to 1e-3.

  • cone_angle (Optional[float]) – Optional. Cone angle for linearly-increased step size. 0. means constant step size. Default: 0.0.

  • traverse_steps_limit (Optional[int]) – Optional. Maximum number of samples per ray.

  • over_allocate (Optional[bool]) – Optional. Whether to over-allocate the memory for the outputs.

  • rays_mask (Optional[Tensor]) – Optional. (n_rays,) Skip some rays if given.

  • t_sorted (Optional[Tensor]) – Optional. (n_rays, n_grids * 2) Pre-computed sorted t values for each ray-grid pair. Default to None.

  • t_indices (Optional[Tensor]) – Optional. (n_rays, n_grids * 2) Pre-computed sorted t indices for each ray-grid pair. Default to None.

  • hits (Optional[Tensor]) – Optional. (n_rays, n_grids) Pre-computed hit flags for each ray-grid pair. Default to None.

  • binaries (Tensor) –

Returns:

A RayIntervals object containing the intervals of the ray traversal, and a RaySamples object containing the samples within each interval. t Tensor of shape (n_rays,) containing the terminated t values for each ray.

Return type:

Tuple[RayIntervals, RaySamples, Tensor]