ResNeSt-50 feature-extracting backbone (no classification head).
Builds a ResNeSt with ResNet-50 topology
([3, 4, 6, 3]), radix = 2, and the canonical
stem_width = 32 deep stem. Approximately 27.5M parameters
and 81.1% ImageNet-1k top-1 accuracy in Zhang et al., 2022
(Table 4) — roughly 5 points higher than plain ResNet-50 at
the same parameter budget. The default ResNeSt configuration.
Model Size
Parameters
pretrainedbool= FalseReserved for future pretrained-weight loading. Currently
ignored.
**overridesobject= {}Keyword overrides forwarded into
ResNeStConfig
(e.g. radix=4 to widen the split-attention).Returns
ResNeStBackbone with the ResNeSt-50 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_50
>>> model = resnest_50()
>>> x = lucid.randn(1, 3, 224, 224)
>>> out = model(x)
>>> out.last_hidden_state.shape
(1, 2048, 7, 7)