Build loss from options. Args: opt (dict): Configuration. It must contain: type (str): Model type.

Initialize the OneHot layer. Parameters ---------- num_classes : int The number of classes to encode.

Split the input tensor (revisits) into two parts/halves. Parameters ---------- x : Tensor The input tensor of sh

Calculate metric from data and options. Args: opt (dict): Configuration. It must contain: type (str): Model type.

Load yaml file or string. Args: f (str): File path or a python string. Returns: dict: Loaded dict.

Initialize network weights. Args: module_list (list[nn.Module] | nn.Module): Modules to be initialized. scale (float): Scale init

Support OrderedDict for yaml. Returns: tuple: yaml Loader and Dumper.

Pad the input tensor with black revisits to a power of 2. Parameters ---------- x : Tensor The input tensor of s

Pixel unshuffle. Args: x (Tensor): Input feature with shape (b, c, hh, hw). scale (int): Downsample ratio. Returns:

Reorder images to 'HWC' order. If the input_order is (h, w), return (h, w, 1); If the input_order is (c, h, w), return (h, w, c); If the

Change to Y channel of YCbCr. Args: img (ndarray): Images with range [0, 255]. Returns: (ndarray): Images with range [0, 255

Initialize the best metric results dict for recording the best metric value and iteration.

Initialize the best metric results dict for recording the best metric value and iteration.

Initialize the best metric results dict for recording the best metric value and iteration.

Convert a BGR image to YCbCr image. The bgr version of rgb2ycbcr. It implements the ITU-R BT.601 conversion for standard-definition telev

Make layers by stacking the same blocks. Args: basic_block (nn.module): nn.module class for basic block. num_basic_block (int): n

Compute the reference frame as the median of all low-res revisits. Parameters ---------- x : Tensor The input te

Initialize the DoubleConv2d layer. Parameters ---------- in_channels : int The number of input channels.

Initialize the ResidualBlock layer. Parameters ---------- in_channels : int The number of input channels.

Initialize the FusionBlock layer. Fuse workflow: xx ---> xx ---> x | ^ |-------^ Parameters -

Initialize the RecursiveFusion layer. Parameters ---------- in_channels : int The number of input channels.

Initialize the ConvTransposeBlock layer. Parameters ---------- in_channels : int The number of input channels.

Initialize the Super-resolution CNN. Parameters ---------- in_channels : int The number of input channels.

Implementation of cPSNR from PROBA-V: https://kelvins.esa.int/proba-v-super-resolution/scoring/ Adds maximization of translations and br

Warp an image or feature map with optical flow. Args: x (Tensor): Tensor with size (n, c, h, w). flow (Tensor): Tensor with size

Forward pass of the High-resolution CNN. Parameters ---------- x : Tensor The input tensor (low-res revisits).

Forward pass. Parameters ---------- x : Tensor The input tensor. Returns ------- Tenso

Forward pass of the DoubleConv2d layer. Parameters ---------- x : Tensor The input tensor of shape (batch_size,

Forward pass of the ResidualBlock layer. Parameters ---------- x : Tensor The input tensor of shape (batch_size,

Forward pass of the DenseBlock layer. Parameters ---------- x : Tensor The input tensor of shape (batch_size, in

Forward pass of the FusionBlock layer. Parameters ---------- x : Tensor The input tensor of shape (batch_size,

Forward pass of the RecursiveFusion layer. Parameters ---------- x : Tensor The input tensor of shape (batch_siz

Forward pass of the ConvTransposeBlock layer. Parameters ---------- x : Tensor The input tensor of shape (batch_

Forward pass of the Super-resolution CNN. Parameters ---------- x : Tensor The input tensor (low-res revisits).