Generate feature. Args: x (Tensor): Input tensor with shape (n, c, h, w). Returns: Tensor: Forward results.

Forward function in generator training step. In this function, we mainly compute the loss items for generator with the given (fake_re

Forward function in generator training step. In this function, we mainly compute the loss items for generator with the given (fake_re

Get the two-dimensional Gaussian filter matrix coefficients. Args: kernel_size (tuple[int]): Kernel filter size in the x and y

Get irregular mask with the constraints in mask ratio. Args: img_shape (tuple[int]): Size of the image. area_ratio_range (tuple(f

Get mean latent of W space in Style-based GANs. Args: generator (nn.Module): Generator of a Style-based GAN. num_samples (int, op

Generate padding function for CAIN. This function produces two functions to pad and depad a tensor, given the number of pixels to be padded.

Get parameters for a random sized crop. Args: data (np.ndarray): Image of type numpy array to be cropped. Returns:

Get the scale factor given input size and output size. Args: input_size (tuple(int)): The size of the input image. output_size (t

Get the output size given input size and scale factor. Args: input_size (tuple): The size of the input image. scale_factor (float

Get weights and indices for interpolation. Args: input_length (int): Length of the input sequence. output_length (int): Length of

Calculate gradient penalty for wgan-gp. Args: discriminator (nn.Module): Network for the discriminator. real_data (Tensor): Real

Init weights for the module.

Init weights for ResidualDenseBlock. Use smaller std for better stability and performance. We empirically use 0.1. See more details i

Initialize the model network weights. Args: pretrained (str, optional): Path to the pretrained weight. Defaults t

Init weights for models. Args: pretrained (str, optional): Path for pretrained weights. If given None, pretrained

Init weights for models. Args: pretrained (str, optional): Path for pretrained weights. If given None, pretrained

Init weights for models. Args: pretrained (str, optional): Path for pretrained weights. If given None, pretrained

Init weights for models. Args: pretrained (str, optional): Path for pretrained weights. If given None, pretrained

Init weights for models. Args: pretrained (str, optional): Path for pretrained weights. If given None, pretrained

Init weights for models. Args: pretrained (str, optional): Path for pretrained weights. If given None, pretrained

Initialize weights for ResidualBlockNoBN. Initialization methods like `kaiming_init` are for VGG-style modules. For modules with resi

Initialize weights for PixelShufflePack.

Initialize weights for the model. Args: pretrained (str, optional): Path for pretrained weights. If given None, p

Initialize weights for the model. Args: pretrained (str, optional): Path for pretrained weights. If given None, p

Init weights for models. Args: pretrained (str, optional): Path for pretrained weights. If given None, pretrained

Init weights. Args: model (nn.Module): Models to be inited. pretrained (str): Path for pretrained weights.

Load annotations for SR dataset. Annotation file is a `npy` file, which contains a list of dict. It loads the GT image path and land

Load annotations for dataset. Returns: list[dict]: Contain dataset annotations.

Load annotations for REDS dataset. Returns: list[dict]: A list of dicts for paired paths and other information.

Load annotations for VimeoK dataset. Returns: list[dict]: A list of dicts for paired paths and other information.

Load annotations for Vid4 dataset. Returns: list[dict]: A list of dicts for paired paths and other information.

Load annotations for VimeoK dataset. Returns: list[dict]: A list of dicts for paired paths and other information.

Load annotations for SR dataset. It loads the LQ and GT image path from folders. Returns: list[dict]: A list of dicts fo

Load annotations for SR dataset. It loads the LQ and GT image path from the annotation file. Each line in the annotation file contain

Load annotations for the dataset. Returns: list[dict]: Returned list of dicts for paired paths of LQ and GT.

Abstract function for loading annotation. All subclasses should overwrite this function

Load annotations for the dataset. Returns: list[dict]: A list of dicts for paired paths and other information.

Load annotations for SR dataset. It loads the LQ and GT image path from the ``meta_info.txt`` in the LMDB files. Returns:

Load annotations for Vimeo-90K dataset. Returns: list[dict]: A list of dicts for paired paths and other information.

Load annotations for SR dataset. It loads the GT image path from folder. Returns: list[dict]: A list of dicts for path o

Load paired image paths. Returns: list[dict]: List that contains paired image paths.

Load annoations for Vimeo-90K dataset. Returns: list[dict]: A list of dicts for paired paths and other information.

Load annoations for the test dataset. Returns: list[dict]: A list of dicts for paired paths and other information.

Load annotations for SR dataset. It loads the ref, LQ and GT image path from folders. Returns: list[dict]: A list of dic

Load annotations for REDS dataset. Returns: list[dict]: A list of dicts for paired paths and other information.

Load Inception Model from given `style` and `inception_kwargs`. This function would try to load Inception under the guidance of `style` given

Add mask weight for correlation map. Add a negative infinity number to the masked regions so that softmax function will result in 'ze

Apply element-wise weight and reduce loss. Args: loss (Tensor): Element-wise loss. weight (Tensor): Element-wise weights. Default

merge input frames and output frames. Interpolate a frame between the given two frames. Merged from [[in1, in2, in3, in4

Compute the Maximum Mean Discrepancy.

Calculate NIQE (Natural Image Quality Evaluator) metric. Ref: Making a "Completely Blind" Image Quality Analyzer. This implementation could p

nn.Module: normalization layer after the first convolution layer

nn.Module: normalization layer after the first convolution layer

nn.Module: normalization layer after the second convolution layer

nn.Module: normalization layer after the second convolution layer

nn.Module: normalization layer after the second convolution layer

Normalize the input image. Args: img (Tensor): Input image. Returns: Tensor: Normalized image.

Converts image in np.array format to PIL image. From C x W x H [0..1] to W x H x C [0...255]

Converts image in np.array format to PIL image. From C x W x H [0..1] to W x H x C [0...255]

Parse information from a line in a requirements text file.

Copy patches using deconv. Args: attention_score (torch.Tensor): Tensor with shape of (n, l , h, w). context_filter (

Calculate patch correlation. Args: x (torch.Tensor): Input tensor. kernel (torch.Tensor): Kernel tensor. Ret

Converts image in PIL format to np.array. From W x H x C [0...255] to C x W x H [0..1]

Converts image in PIL format to np.array. From W x H x C [0...255] to C x W x H [0..1]

Pixel Normalization. This normalization is proposed in: Progressive Growing of GANs for Improved Quality, Stability, and Variation Args:

Prepare testing data. Args: idx (int): Index for getting each testing batch. Returns: Tensor: Returned testi

Prepare training data. Args: idx (int): Index of the training batch data. Returns: dict: Returned training b