mobilenet_v3_small(pretrained: bool = False, overrides: object = {})MobileNet-v3-Small feature-extracting backbone.
Builds a MobileNetV3 with the NAS-designed Small
topology from Howard et al., 2019 (Table 2): 11 inverted-residual
bottleneck blocks with selective SE attention and aggressive
hard-swish usage from the very first stage, followed by a 1×1
expansion to 576 channels. Approximately 2.9M parameters and
67.4% ImageNet-1k top-1 accuracy (Table 3) — the lower-latency
variant of the family, targeting tight mobile budgets.
Model Size
Parameters
pretrainedbool= FalseReserved for future pretrained-weight loading. Currently
ignored.
**overridesobject= {}Keyword overrides forwarded into
MobileNetV3Config.Returns
MobileNetV3Backbone with the MobileNet-v3-Small configuration applied
(or with overrides merged on top of it).
Notes
See Howard et al., "Searching for MobileNetV3", ICCV 2019 (arXiv:1905.02244), Table 2 (V3-Small spec) and Table 3 (ImageNet accuracy).
Examples
>>> import lucid
>>> from lucid.models.vision.mobilenet_v3 import mobilenet_v3_small
>>> model = mobilenet_v3_small()
>>> x = lucid.randn(1, 3, 224, 224)
>>> out = model(x)
>>> out.last_hidden_state.shape
(1, 576, 1, 1)