VGG-11 feature-extracting backbone (config A, no BatchNorm).
Builds a VGG with the paper-cited VGG-11 topology:
per-block convolution counts (1, 1, 2, 2, 2) across the five
blocks of widths (64, 128, 256, 512, 512). Approximately
132.9 M parameters in the full classifier variant (≈9.2 M in the
convolutional trunk alone). Reaches a top-5 ImageNet validation
error of 9.4% in Simonyan & Zisserman, ICLR 2015 (Table 3,
configuration A).
Model Size
Parameters
pretrainedbool= FalseReserved for future pretrained-weight loading. Currently
ignored — the returned model is randomly initialised.
**overridesobject= {}Keyword overrides forwarded into
VGGConfig to
customise individual fields (in_channels, num_classes,
dropout).Returns
VGGBackbone with the VGG-11 configuration applied (or with
overrides merged on top of it).
Notes
See Simonyan & Zisserman, "Very Deep Convolutional Networks for Large-Scale Image Recognition", ICLR 2015, Table 1 (configuration A).
Examples
>>> import lucid
>>> from lucid.models.vision.vgg import vgg_11
>>> model = vgg_11()
>>> x = lucid.randn(1, 3, 224, 224)
>>> out = model(x)
>>> out.last_hidden_state.shape # (B, 512, 7, 7)
(1, 512, 7, 7)