ResNet-200 feature-extracting backbone (no classification head).
Builds a ResNet with the deep bottleneck topology
introduced in He et al., "Identity Mappings in Deep Residual
Networks", ECCV 2016 (arXiv:1603.05027): _Bottleneck
blocks stacked [3, 24, 36, 3] over four stages — approximately
64.7M parameters. Used as a high-capacity backbone for ImageNet
pre-training in large-scale vision pipelines.
Model Size
Parameters
pretrainedbool= FalseReserved for future pretrained-weight loading. Currently
ignored.
**overridesobject= {}Keyword overrides forwarded into
ResNetConfig.Returns
ResNetBackbone with the ResNet-200 configuration applied (or with
overrides merged on top of it).
Notes
Final-stage output is 2048 channels. Memory cost scales with the
24-block stage-2 — keep zero_init_residual=True for stable
convergence at large batch sizes.
Examples
>>> import lucid
>>> from lucid.models.vision.resnet import resnet_200
>>> model = resnet_200()
>>> x = lucid.randn(1, 3, 224, 224)
>>> out = model(x)
>>> out.last_hidden_state.shape # (B, 2048, H/32, W/32)
(1, 2048, 7, 7)