`mmagic.evaluation.functional.fid_inception`¶

Inception networks used in calculating FID and Inception metrics.

This code is modified from: https://github.com/rosinality/stylegan2-pytorch/blob/master/inception.py

Module Contents¶

Classes¶

`InceptionV3`	Pretrained InceptionV3 network returning feature maps.
`FIDInceptionA`	InceptionA block patched for FID computation.
`FIDInceptionC`	InceptionC block patched for FID computation.
`FIDInceptionE_1`	First InceptionE block patched for FID computation.
`FIDInceptionE_2`	Second InceptionE block patched for FID computation.

Functions¶

fid_inception_v3([load_ckpt])

Build pretrained Inception model for FID computation.

Attributes¶

FID_WEIGHTS_URL

mmagic.evaluation.functional.fid_inception.FID_WEIGHTS_URL = https://github.com/mseitzer/pytorch-fid/releases/download/fid_weights/pt_inception-2015-12-05-672...[source]¶

class mmagic.evaluation.functional.fid_inception.InceptionV3(output_blocks=[DEFAULT_BLOCK_INDEX], resize_input=True, normalize_input=True, requires_grad=False, use_fid_inception=True, load_fid_inception=True)[source]¶

Bases: torch.nn.Module

Pretrained InceptionV3 network returning feature maps.

DEFAULT_BLOCK_INDEX = 3[source]¶

BLOCK_INDEX_BY_DIM[source]¶

forward(inp)[source]¶

Get Inception feature maps.

Parameters: inp (torch.Tensor) – Input tensor of shape Bx3xHxW. Values are expected to be in range (0, 1)
Returns: Corresponding to the selected output block, sorted ascending by index.
Return type: list(torch.Tensor)

mmagic.evaluation.functional.fid_inception.fid_inception_v3(load_ckpt=True)[source]¶

Build pretrained Inception model for FID computation.

The Inception model for FID computation uses a different set of weights and has a slightly different structure than torchvision’s Inception.

This method first constructs torchvision’s Inception and then patches the necessary parts that are different in the FID Inception model.

class mmagic.evaluation.functional.fid_inception.FIDInceptionA(in_channels, pool_features)[source]¶

Bases: torchvision.models.inception.InceptionA

InceptionA block patched for FID computation.

forward(x)[source]¶

Get InceptionA feature maps.

Parameters: x (torch.Tensor) – Input tensor of shape BxCxHxW.
Returns: Feature Maps of x outputted by this block.
Return type: torch.Tensor

class mmagic.evaluation.functional.fid_inception.FIDInceptionC(in_channels, channels_7x7)[source]¶

Bases: torchvision.models.inception.InceptionC

InceptionC block patched for FID computation.

forward(x)[source]¶

Get InceptionC feature maps.

Parameters: x (torch.Tensor) – Input tensor of shape BxCxHxW.
Returns: Feature Maps of x outputted by this block.
Return type: torch.Tensor

class mmagic.evaluation.functional.fid_inception.FIDInceptionE_1(in_channels)[source]¶

Bases: torchvision.models.inception.InceptionE

First InceptionE block patched for FID computation.

forward(x)[source]¶

Get first InceptionE feature maps.

Parameters: x (torch.Tensor) – Input tensor of shape BxCxHxW.
Returns: Feature Maps of x outputted by this block.
Return type: torch.Tensor

class mmagic.evaluation.functional.fid_inception.FIDInceptionE_2(in_channels)[source]¶

Bases: torchvision.models.inception.InceptionE

Second InceptionE block patched for FID computation.

forward(x)[source]¶

Get second InceptionE feature maps.

Parameters: x (torch.Tensor) – Input tensor of shape BxCxHxW.
Returns: Feature Maps of x outputted by this block.
Return type: torch.Tensor

mmagic.evaluation.functional.fid_inception¶

Module Contents¶

Classes¶

Functions¶

Attributes¶

`mmagic.evaluation.functional.fid_inception`¶