An attribute dictionary that automatically handles nested keys joined by "/". Originally copied from: https://stackoverflow.com/questions/30

An array of points sampled on a surface. Each point may have zero or more channel attributes. :param coords: an [N x 3] array of point c

A Camera implementation for a standard pinhole camera. The camera rays shoot away from the origin in the z direction, with the x and y d

Instead of fixed intervals, a sample is drawn uniformly at random from each interval.

Implements a batch, differentiable, standard pinhole camera

Only does cross attention

A ray in ray casting.

Interact with a dataset zipfile exported by view_data.py.

Utilities for training and sampling diffusion models. Ported directly from here: https://github.com/hojonathanho/diffusion/blob/1e0dceb3

The result of casting N rays onto a mesh.

Stores the relevant state and results of integrate( lambda t: density(t) * channels(t) * transmittance(t), [t0,

A 3D triangle mesh with optional data at the vertices and faces.

Adapted from here, but effectively shines the light in both positive and negative directions: https://github.com/facebookresearch/pytorch3d/b

Annotate a differentiable camera with a multi-dimensional batch shape.

A wrapper around a pre-trained CLIP model that automatically handles batches of texts, images, and embeddings.

Given the initial estimate of densities, this samples more from regions/bins expected to have objects.

Encode cameras and views using a transformer model with extra output token(s) used to extract a latent vector.

Renders rays using stratified sampling only unlike vanilla NeRF. The same setup as NeRF++.

Encode point clouds using a transformer model with an extra output token used to extract a latent vector.

Encode point clouds using a transformer model with an extra output token used to extract a latent vector.

Encode point clouds using a transformer model with an extra output token used to extract a latent vector.

Encode point clouds using a transformer model with an extra output token used to extract a latent vector.

A diffusion process which can skip steps in a base diffusion process. :param use_timesteps: (unordered) timesteps from the original diffusion

A 3D triangle mesh with optional data at the vertices and faces.

Coarse and fine-grained rendering as proposed by NeRF. This class additionally supports background rendering like NeRF++.

Axis-aligned bounding box defined by the two opposite corners.

An object describing how a camera corresponds to pixels in an image.

An object describing how a camera corresponds to pixels in an image.

A ViewData that is implemented in memory.

Base class for PyTorch meta-learning modules. These modules accept an additional argument `params` in their `forward` method. Notes

Parametric scene representation whose outputs are integrated by NeRFRenderer

Encode point clouds using a perceiver model with an extra output token used to extract a latent vector.

Encode point clouds using a perceiver model with an extra output token used to extract a latent vector.