ResNeSt-14 feature-extracting backbone (no classification head).
Builds a ResNeSt with block repeats [1, 1, 1, 1]
over four stages — a lightweight variant of the ResNeSt family
targeted at low-latency deployments. Uses radix = 2 and
stem_width = 32. Approximately 10.6M parameters.
Model Size
Parameters
pretrainedbool= FalseReserved for future pretrained-weight loading. Currently
ignored — the returned model is randomly initialised.
**overridesobject= {}Keyword overrides forwarded into
ResNeStConfig.Returns
ResNeStBackbone with the ResNeSt-14 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). Final-stage output is 2048 channels.
Examples
>>> import lucid
>>> from lucid.models.vision.resnest import resnest_14
>>> model = resnest_14()
>>> x = lucid.randn(1, 3, 224, 224)
>>> out = model(x)
>>> out.last_hidden_state.shape
(1, 2048, 7, 7)