ResNeSt-101 feature-extracting backbone (no classification head).
Builds a ResNeSt with ResNet-101 topology
([3, 4, 23, 3]), radix = 2, and a wider
stem_width = 64 deep stem. Approximately 48.3M parameters
and 82.8% ImageNet-1k top-1 accuracy in Zhang et al., 2022
(Table 4). Higher-accuracy drop-in replacement for ResNeSt-50.
Model Size
Parameters
pretrainedbool= FalseReserved for future pretrained-weight loading. Currently
ignored.
**overridesobject= {}Keyword overrides forwarded into
ResNeStConfig.Returns
ResNeStBackbone with the ResNeSt-101 configuration applied (or
with overrides merged on top of it).
Notes
See Zhang et al., "ResNeSt: Split-Attention Networks", CVPR Workshops 2022 (arXiv:2004.08955), Table 4.
Examples
>>> import lucid
>>> from lucid.models.vision.resnest import resnest_101
>>> model = resnest_101()
>>> x = lucid.randn(1, 3, 224, 224)
>>> out = model(x)
>>> out.last_hidden_state.shape
(1, 2048, 7, 7)