mmedit.models.editors.stylegan1.stylegan1_modules

Module Contents

Classes

EqualLinearActModule	Equalized LR Linear Module with Activation Layer.
NoiseInjection	Noise Injection Module.
ConstantInput	Constant Input.
Blur	Blur module.
AdaptiveInstanceNorm	Adaptive Instance Normalization Module.
StyleConv	Base class for all neural network modules.

Functions

make_kernel(k)

class mmedit.models.editors.stylegan1.stylegan1_modules.EqualLinearActModule(*args, equalized_lr_cfg=dict(gain=1.0, lr_mul=1.0), bias=True, bias_init=0.0, act_cfg=None, **kwargs)

Bases: torch.nn.Module

Equalized LR Linear Module with Activation Layer.

This module is modified from EqualizedLRLinearModule defined in PGGAN. The major features updated in this module is adding support for activation layers used in StyleGAN2.

参数

• equalized_lr_cfg (dict | None, optional) – Config for equalized lr. Defaults to dict(gain=1., lr_mul=1.).

• bias (bool, optional) – Whether to use bias item. Defaults to True.

• bias_init (float, optional) – The value for bias initialization. Defaults to 0..

• act_cfg (dict | None, optional) – Config for activation layer. Defaults to None.

forward(x)

Forward function.

参数

x (Tensor) – Input feature map with shape of (N, C, …).

返回

Output feature map.

返回类型

Tensor

class mmedit.models.editors.stylegan1.stylegan1_modules.NoiseInjection(noise_weight_init=0.0)

Bases: torch.nn.Module

Noise Injection Module.

In StyleGAN2, they adopt this module to inject spatial random noise map in the generators.

参数

noise_weight_init (float, optional) – Initialization weight for noise injection. Defaults to 0..

forward(image, noise=None, return_noise=False)

Forward Function.

参数

• image (Tensor) – Spatial features with a shape of (N, C, H, W).

• noise (Tensor, optional) – Noises from the outside. Defaults to None.

• return_noise (bool, optional) – Whether to return noise tensor. Defaults to False.

返回

Output features.

返回类型

Tensor

class mmedit.models.editors.stylegan1.stylegan1_modules.ConstantInput(channel, size=4)

Bases: torch.nn.Module

Constant Input.

In StyleGAN2, they substitute the original head noise input with such a constant input module.

参数

• channel (int) – Channels for the constant input tensor.

• size (int, optional) – Spatial size for the constant input. Defaults to 4.

forward(x)

Forward function.

参数

x (Tensor) – Input feature map with shape of (N, C, …).

返回

Output feature map.

返回类型

Tensor

mmedit.models.editors.stylegan1.stylegan1_modules.make_kernel(k)

class mmedit.models.editors.stylegan1.stylegan1_modules.Blur(kernel, pad, upsample_factor=1)

Bases: torch.nn.Module

Blur module.

This module is adopted rightly after upsampling operation in StyleGAN2.

参数

• kernel (Array) – Blur kernel/filter used in UpFIRDn.

• pad (list[int]) – Padding for features.

• upsample_factor (int, optional) – Upsampling factor. Defaults to 1.

forward(x)

Forward function.

参数

x (Tensor) – Input feature map with shape of (N, C, H, W).

返回

Output feature map.

返回类型

Tensor

class mmedit.models.editors.stylegan1.stylegan1_modules.AdaptiveInstanceNorm(in_channel, style_dim)

Bases: torch.nn.Module

Adaptive Instance Normalization Module.

Ref: https://github.com/rosinality/style-based-gan-pytorch/blob/master/model.py # noqa

参数

• in_channel (int) – The number of input’s channel.

• style_dim (int) – Style latent dimension.

forward(input, style)

Forward function.

参数

• input (Tensor) – Input tensor with shape (n, c, h, w).

• style (Tensor) – Input style tensor with shape (n, c).

返回

Forward results.

返回类型

Tensor

class mmedit.models.editors.stylegan1.stylegan1_modules.StyleConv(in_channels, out_channels, kernel_size, style_channels, padding=1, initial=False, blur_kernel=[1, 2, 1], upsample=False, fused=False)

Bases: torch.nn.Module

Base class for all neural network modules.

Your models should also subclass this class.

Modules can also contain other Modules, allowing to nest them in a tree structure. You can assign the submodules as regular attributes:

import torch.nn as nn
import torch.nn.functional as F

class Model(nn.Module):
    def __init__(self):
        super().__init__()
        self.conv1 = nn.Conv2d(1, 20, 5)
        self.conv2 = nn.Conv2d(20, 20, 5)

    def forward(self, x):
        x = F.relu(self.conv1(x))
        return F.relu(self.conv2(x))

Submodules assigned in this way will be registered, and will have their parameters converted too when you call to(), etc.

备注

As per the example above, an __init__() call to the parent class must be made before assignment on the child.

变量

training (bool) – Boolean represents whether this module is in training or evaluation mode.

forward(x, style1, style2, noise1=None, noise2=None, return_noise=False)

Forward function.

参数

• x (Tensor) – Input tensor.

• style1 (Tensor) – Input style tensor with shape (n, c).

• style2 (Tensor) – Input style tensor with shape (n, c).

• noise1 (Tensor, optional) – Noise tensor with shape (n, c, h, w). Defaults to None.

• noise2 (Tensor, optional) – Noise tensor with shape (n, c, h, w). Defaults to None.

• return_noise (bool, optional) – If True, noise1 and noise2

• False. (will be returned with out. Defaults to) –

返回

Forward results.

返回类型

Tensor | tuple[Tensor]