2D convolution class Arg(s): in_channels : int number of input channels out_channels : int number of

2D convolution class Arg(s): in_channels : int number of input channels out_channels : int number of

2D convolution class Arg(s): in_channels : int number of input channels out_channels : int number of

ScaffNet to infer topology from sparse points Arg(s): encoder_type : list[str] encoder types: vggnet08, vggnet11, vggnet

Computes and stores the average and current value

Wrapper class for all external depth completion models Arg(s): model_name : str depth completion model to use mi

Network to fuse image and depth information together and learns residual for initial depth Arg(s): encoder_type : list[str]

Apply a user-defined lambda as a transform. Args: lambd (function): Lambda/function to be used for transform.

Class to perform outlier removal based on depth difference in local neighborhood Arg(s): kernel_size : int local neighbo

Class to perform outlier removal based on depth difference in local neighborhood Arg(s): kernel_size : int local neighbo

2D atrous convolution class Arg(s): in_channels : int number of input channels out_channels : int nu

A data loader for the Kitti dataset

Calibrated Backprojection Network class Arg(s): input_channels_image : int number of channels in the image input

Monitored distillation student model class Arg(s): encoder_type : list[str] encoder type (ex: ['kbnet', 'batch_norm'])

Class to perform outlier removal based on depth difference in local neighborhood Arg(s): kernel_size : int local neighbo

Pose network for computing relative pose between a pair of images Arg(s): encoder_type : str posenet, resnet18, resnet34

ResNet encoder with skip connections Arg(s): n_layer : int architecture type based on layers: 18, 34, 50 input_c

VGGNet block class Arg(s): in_channels : int number of input channels out_channels : int number of o

VGGNet block class Arg(s): in_channels : int number of input channels out_channels : int number of o

Composes several transforms together. Args: transforms (list of ``Transform`` objects): list of transforms to compose. Example:

Class for interfacing with ENet model Arg(s): device : torch.device device to run model on max_depth : float

Class for interfacing with FusionNet model Arg(s): dataset_name : str model for a given dataset min_predict_dept

Class for interfacing with KBNet model Arg(s): dataset_name : str model for a given dataset min_predict_depth :

Class for interfacing with MSGCHN model Arg(s): device : torch.device device to run model on max_depth : float

Class for interfacing with NLSPN model Arg(s): device : torch.device device to run model on max_depth : float

Class for interfacing with PENet model Arg(s): device : torch.device device to run model on max_depth : float

Pose network for computing relative pose between a pair of images Arg(s): encoder_type : str posenet, resnet18, resnet34

Pose network for computing relative pose between a pair of images Arg(s): encoder_type : str posenet, resnet18, resnet34

Class for interfacing with ScaffNet model Arg(s): dataset_name : str model for a given dataset min_predict_depth

Transpose convolution class Arg(s): in_channels : int number of input channels out_channels : int nu

Transpose convolution class Arg(s): in_channels : int number of input channels out_channels : int nu

Transpose convolution class Arg(s): in_channels : int number of input channels out_channels : int nu

Up-convolution (upsample + convolution) block class Arg(s): in_channels : int number of input channels out_chann

Up-convolution (upsample + convolution) block class Arg(s): in_channels : int number of input channels out_chann

Up-convolution (upsample + convolution) block class Arg(s): in_channels : int number of input channels out_chann

Add coords to a tensor

Basic atrous ResNet block class Arg(s): in_channels : int number of input channels out_channels : int

ResNet encoder with skip connections Arg(s): n_layer : int architecture type based on layers: 18, 34 input_chann

Atrous Spatial Pyramid Pooling class Arg(s): in_channels : int number of input channels out_channels : int

Atrous VGGNet block class (last block performs atrous convolution instead of convolution with stride) Arg(s): in_channels : int

Atrous VGGNet encoder with skip connections Arg(s): input_channels : int number of channels in input data n_laye

Crops the given ``numpy.ndarray`` at the bottom. Args: size (sequence or int): Desired output size of the crop. If size is an

Calibrated backprojection (KB) layer class Arg(s): in_channels_image : int number of input channels for image (RGB) bran

Calibrated backprojection (KB) layer class Arg(s): in_channels_image : int number of input channels for image (RGB) bran

Crops the given ``numpy.ndarray`` at the center. Args: size (sequence or int): Desired output size of the crop. If size is an

Randomly change the brightness, contrast and saturation of an image. Args: brightness (float): How much to jitter brightness. brightness_

Crops the given ``numpy.ndarray`` at the center. Args: size (sequence or int): Desired output size of the crop. If size is an

Decoder block with skip connections (U-Net) Arg(s): in_channels : int number of input channels skip_channels : i

Decoder block with skip connection Arg(s): in_channels : int number of input channels skip_channels : int

Decoder block with skip connection Arg(s): in_channels : int number of input channels skip_channels : int

Dataset for fetching: (1) image (2) sparse depth (3) intrinsic camera calibration matrix Arg(s): image_paths

Depthwise separable convolution class Performs 1. separate k x k convolution per channel (depth-wise) 2. 1 x 1 convolution across all

Depthwise separable convolution class Performs 1. separate k x k convolution per channel (depth-wise) 2. 1 x 1 convolution across all

Fully connected layer Arg(s): in_channels : int number of input neurons out_channels : int number of

Dataset for fetching: (1) image (2) sparse depth Arg(s): image_paths : list[str] paths to image triplets

Dataset for fetching: (1) image (2) sparse depth Arg(s): image_paths : list[str] paths to image triplets

Dataset for fetching: (1) image at time t-1, t, and t+1 (2) sparse depth at time t (3) camera intrinsics matrix Arg(

Dataset for fetching: (1) image at time t-1, t, and t+1 (2) dense depth at time t (3) sparse depth at time t (4)

Horizontally flip the given ``numpy.ndarray``. Args: do_flip (boolean): whether or not do horizontal flip.

KBNet decoder with skip connections Arg(s): input_channels : int number of channels in input latent vector outpu

Calibrated backprojection network (KBNet) encoder with skip connections Arg(s): in_channels_image : int number of input

Calibrated backprojection network (KBNet) encoder with skip connections Arg(s): in_channels_image : int number of input

Dataset for fetching: (1) image (2) sparse depth (3) camera intrinsics matrix Arg(s): image_paths : list[str

Dataset for fetching: (1) image at time t-1, t, and t+1 (2) sparse depth (3) camera intrinsics matrix Arg(s):