r"""Get attribute.

Alpha compositing of (sampled) MPIs given their RGBs and alphas. Args: alphas (tensor [batch,ray,samples]): The predicted opacity values.

r"""Normalize using ImageNet mean and std. Args: input (4D tensor NxCxHxW): The input images, assuming to be [-1, 1]. Returns:

r"""Apply any ops from op_dict to data types. Args: data (dict): Dict containing data_type as key, with each value

r"""Calibrate batch norm momentum Args: m: Pytorch module

Composite the samples to render the RGB/depth/opacity of the corresponding pixels. Args: quantities (tensor [batch,ray,samples,k]): The qu

Contraction function on mip-NeRF 360 Gaussians. Args: mean (tensor [...,3]): The mean values. cov (tensor [...,3,3]): The covaria

r"""Decode list of strings in json to objects.

r"""Decode list of pickled strings to objects.

r"""Get number of devices

r""" Reduce to rank 0

r"""Flip hdr images. Args: data (dict): Keys are from data types. Values can be numpy.ndarray or list of numpy.nd

r"""Write the value in the tensorboard. Args: step (int): Epoch or iteration number.

r"""PyTorch module forward function overload.

r"""PyTorch module forward function overload.

Compute Fundamental Matrix from a relative pose and intrinsics. param poses: [views, 3, 4] tensor representing relative poses. intr1: [

r"""Get attribute with default choice. If attribute does not exist, set it using the default value. Args: cfg (obj) : Config options.

Args: pose (tensor [3,4]/[B,3,4]): Camera pose. intr (tensor [3,3]/[B,3,3]): Camera intrinsics. image_size (list of int):

r"""Get obect name

r"""Iteratively try to get the attribute from cfg. If not found, return default. Args: cfg (obj): Config file. attr_name (str

Compute the intersection of a ray with a depth plane. Args: center (tensor [B,HW,3]): Camera center of the target pose. ray (tenso

r"""Get weight state Args: mod: Pytorch module

r"""Return the gradient norm of model. Args: model (PyTorch module): Your network.

r"""Init function Args: m: module to be weight initialized.

Interpolate between two poses with Slerp. Args: pose_a (tensor [...,3,4]): Pose at time t=0. pose_b (tensor [...,3,4])

r"""L2 normalization.

r"""Convert list of numeric values to numpy array.

r"""Convert list of numeric values to tensor.

r"""Apply this function only to the master GPU.

r"""master-only print

r"""Perform data augmentation on images only. Args: data (dict): Keys are from data types. Values can be numpy.ndarray

Encode input into position codes. Args: input (tensor [bs, ..., N]): A batch of data with N dimension. num_freq_bases: (int): The

r"""Randomly shift the input tensor. Args: x (4D tensor): The input batch of images. offset (int): The maximum offset ratio that

Yield sampled rays for coordinate-based model to predict NeRF. Args: pose (tensor [bs,3,4]): Camera poses ([R,t]). intr (tensor [b

Reparametrize the sampled distance values according to param_type. Args: dist (tensor): Sampled distance values. param_type (str):

r"""Reset batch norm statistics Args: m: Pytorch module

Recursively find object and set value

Sample points on ray shooting from pixels using distance. Args: ray_size (int [2]): Integers for [batch size, number of rays]. ran

Sample points on ray shooting from pixels using the weights from the coarse NeRF. Args: bin (tensor [batch_size, num_rays, intvs]): bins o

Sample n points only the northern view hemisphere within the angle to x.

Scale the camera center from the origin. 0 = R@c+t --> c = -R^T@t (camera center in world coordinates) 0 = R@(sc)+t' --> t' = -R@(sc)

r"""Set summary writer Args: log_dir (str): Log directory.

r"""Slice all tensors from start to end. Args: data (dict, list or tensor): Input data.

r"""Split concatenated labels into their parts. Args: labels (torch.Tensor): Labels obtained through concatenation. label_lengths

r"""Get Driver Version

r"""Compute results images and save the results in the specified folder. Args: data_loader (torch.utils.data.DataLoader): PyTorch

r"""Move all halfs to float. Args: data (dict, list or tensor): Input data.