423A35C7/pytorch3d
1
0
Fork 0
mirror of https://github.com/facebookresearch/pytorch3d.git synced 2025-08-02 20:02:49 +08:00
Code Issues Packages Projects Releases Wiki Activity

raysampling stratified_sampling doc

Browse Source 
Summary: Followup to D39259775 (438c194ec6)

Reviewed By: davnov134

Differential Revision: D39271753

fbshipit-source-id: 5cf11c1210369e1762ef0d5d0c7b60336711b261
This commit is contained in:
Jeremy Reizenstein 2022-09-06 10:28:45 -07:00 committed by Facebook GitHub Bot
parent 73ea4187de
commit df36223ddf
1 changed files with 9 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
pytorch3d/renderer/implicit/raysampling.py
View File

@ -89,7 +89,7 @@ class MultinomialRaysampler(torch.nn.Module):
max_depth: The maximum depth of a ray-point. max_depth: The maximum depth of a ray-point.
n_rays_per_image: If given, this amount of rays are sampled from the grid. n_rays_per_image: If given, this amount of rays are sampled from the grid.
unit_directions: whether to normalize direction vectors in ray bundle. unit_directions: whether to normalize direction vectors in ray bundle.
stratified_sampling: if set, performs stratified random sampling stratified_sampling: if True, performs stratified random sampling
along the ray; otherwise takes ray points at deterministic offsets. along the ray; otherwise takes ray points at deterministic offsets.
""" """
super().__init__() super().__init__()
@ -136,9 +136,8 @@ class MultinomialRaysampler(torch.nn.Module):
max_depth: The maximum depth of a ray-point. max_depth: The maximum depth of a ray-point.
n_rays_per_image: If given, this amount of rays are sampled from the grid. n_rays_per_image: If given, this amount of rays are sampled from the grid.
n_pts_per_ray: The number of points sampled along each ray. n_pts_per_ray: The number of points sampled along each ray.
stratified_sampling: if set, performs stratified sampling in n_pts_per_ray stratified_sampling: if set, overrides stratified_sampling provided
bins for each ray; otherwise takes n_pts_per_ray deterministic points in __init__.
on each ray with uniform offsets.
Returns: Returns:
A named tuple RayBundle with the following fields: A named tuple RayBundle with the following fields:
origins: A tensor of shape origins: A tensor of shape
@ -296,7 +295,7 @@ class MonteCarloRaysampler(torch.nn.Module):
min_depth: The minimum depth of each ray-point. min_depth: The minimum depth of each ray-point.
max_depth: The maximum depth of each ray-point. max_depth: The maximum depth of each ray-point.
unit_directions: whether to normalize direction vectors in ray bundle. unit_directions: whether to normalize direction vectors in ray bundle.
stratified_sampling: if set, performs stratified sampling in n_pts_per_ray stratified_sampling: if True, performs stratified sampling in n_pts_per_ray
bins for each ray; otherwise takes n_pts_per_ray deterministic points bins for each ray; otherwise takes n_pts_per_ray deterministic points
on each ray with uniform offsets. on each ray with uniform offsets.
""" """
@ -322,9 +321,9 @@ class MonteCarloRaysampler(torch.nn.Module):
""" """
Args: Args:
cameras: A batch of `batch_size` cameras from which the rays are emitted. cameras: A batch of `batch_size` cameras from which the rays are emitted.
stratified_sampling: if set, performs stratified sampling in n_pts_per_ray stratified_sampling: if set, overrides stratified_sampling provided
bins for each ray; otherwise takes n_pts_per_ray deterministic points in __init__.
on each ray with uniform offsets.
Returns: Returns:
A named tuple RayBundle with the following fields: A named tuple RayBundle with the following fields:
origins: A tensor of shape origins: A tensor of shape
@ -514,7 +513,7 @@ def _xy_to_ray_bundle(
max_depth: The maximum depth of each ray-point. max_depth: The maximum depth of each ray-point.
n_pts_per_ray: The number of points sampled along each ray. n_pts_per_ray: The number of points sampled along each ray.
unit_directions: whether to normalize direction vectors in ray bundle. unit_directions: whether to normalize direction vectors in ray bundle.
stratified_sampling: if set, performs stratified sampling in n_pts_per_ray stratified_sampling: if True, performs stratified sampling in n_pts_per_ray
bins for each ray; otherwise takes n_pts_per_ray deterministic points bins for each ray; otherwise takes n_pts_per_ray deterministic points
on each ray with uniform offsets. on each ray with uniform offsets.
""" """
@ -583,7 +582,7 @@ def _jiggle_within_stratas(bin_centers: torch.Tensor) -> torch.Tensor:
More specifically, it replaces each point's value `z` More specifically, it replaces each point's value `z`
with a sample from a uniform random distribution on with a sample from a uniform random distribution on
`[z - delta_−, z + delta_+]`, where `delta_−` is half of the difference `[z - delta_-, z + delta_+]`, where `delta_-` is half of the difference
between `z` and the previous point, and `delta_+` is half of the difference between `z` and the previous point, and `delta_+` is half of the difference
between the next point and `z`. For the first and last items, the between the next point and `z`. For the first and last items, the
corresponding boundary deltas are assumed zero. corresponding boundary deltas are assumed zero.