mmagic.models.editors.swinir.swinir_modules

Module Contents

Classes

PatchEmbed	Image to Patch Embedding
PatchUnEmbed	Image to Patch Unembedding
Upsample	Upsample module.
UpsampleOneStep	UpsampleOneStep module (the difference with Upsample is that it always

class mmagic.models.editors.swinir.swinir_modules.PatchEmbed(img_size=224, patch_size=4, in_chans=3, embed_dim=96, norm_layer=None)

Bases: torch.nn.Module

Image to Patch Embedding :param img_size: Image size. Default: 224. :type img_size: int :param patch_size: Patch token size. Default: 4. :type patch_size: int :param in_chans: Number of input image channels. Default: 3. :type in_chans: int :param embed_dim: Number of linear projection output channels.

Default: 96.

Parameters

norm_layer (nn.Module, optional) – Normalization layer. Default: None

forward(x)

Forward function.

Parameters

x (Tensor) – Input tensor with shape (B, C, Ph, Pw).

Returns

Forward results.

Return type

Tensor

class mmagic.models.editors.swinir.swinir_modules.PatchUnEmbed(img_size=224, patch_size=4, in_chans=3, embed_dim=96, norm_layer=None)

Bases: torch.nn.Module

Image to Patch Unembedding :param img_size: Image size. Default: 224. :type img_size: int :param patch_size: Patch token size. Default: 4. :type patch_size: int :param in_chans: Number of input image channels. Default: 3. :type in_chans: int :param embed_dim: Number of linear projection output channels.

Default: 96.

Parameters

norm_layer (nn.Module, optional) – Normalization layer. Default: None

forward(x, x_size)

Forward function.

Parameters

• x (Tensor) – Input tensor with shape (B, L, C).

• x_size (tuple[int]) – Resolution of input feature.

Returns

Forward results.

Return type

Tensor

class mmagic.models.editors.swinir.swinir_modules.Upsample(scale, num_feat)

Bases: torch.nn.Sequential

Upsample module.

Parameters

• scale (int) – Scale factor. Supported scales: 2^n and 3.

• num_feat (int) – Channel number of intermediate features.

class mmagic.models.editors.swinir.swinir_modules.UpsampleOneStep(scale, num_feat, num_out_ch, input_resolution=None)

Bases: torch.nn.Sequential

UpsampleOneStep module (the difference with Upsample is that it always only has 1conv + 1pixelshuffle) Used in lightweight SR to save parameters.

Parameters

• scale (int) – Scale factor. Supported scales: 2^n and 3.

• num_feat (int) – Channel number of intermediate features.

• num_out_ch (int) – Channel number for PixelShuffle.

• input_resolution (tuple[int], optional) – Input resolution. Default: None