SE-ResNet-18 feature-extracting backbone (no classification head).
Builds an SENet with ResNet-18 topology
(_SEBasicBlock blocks stacked [2, 2, 2, 2]) plus
a squeeze-and-excitation module at the end of every block.
Approximately 11.8M parameters — about 1% more than plain
ResNet-18 — at a typical 1.0–1.5 point ImageNet top-1
accuracy improvement.
Model Size
Parameters
pretrainedbool= FalseReserved for future pretrained-weight loading. Currently
ignored — the returned model is randomly initialised.
**overridesobject= {}Keyword overrides forwarded into
SENetConfig
(e.g. reduction=8 to tighten the SE bottleneck).Returns
SENetBackbone with the SE-ResNet-18 configuration applied (or
with overrides merged on top of it).
Notes
See Hu et al., "Squeeze-and-Excitation Networks", CVPR 2018 (arXiv:1709.01507), Table 2.
Examples
>>> import lucid
>>> from lucid.models.vision.senet import se_resnet_18
>>> model = se_resnet_18()
>>> x = lucid.randn(1, 3, 224, 224)
>>> out = model(x)
>>> out.last_hidden_state.shape
(1, 512, 7, 7)