MobileNet-v1 backbone at width multiplier .
Builds a MobileNet with the canonical paper topology:
a 3×3 stem (stride 2) followed by 13 depthwise+pointwise blocks,
yielding approximately 4.2M parameters. Howard et al., 2017
report a 70.6% ImageNet-1k top-1 validation accuracy with this
configuration (Table 6). The default choice when the full
accuracy budget is available.
Model Size
Parameters
pretrainedbool= FalseNo checkpoint is published for
mobilenet, so True raises
NotImplementedError rather than return randomly initialised weights.
mobilenet_cls is the checkpointed variant.**overridesobject= {}Keyword overrides forwarded into
MobileNetConfig
(e.g. in_channels=1 for grayscale input).Returns
MobileNetBackbone with the MobileNet-v1 ()
configuration applied (or with overrides merged on top
of it).
Notes
See Howard et al., "MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications", arXiv:1704.04861, 2017, Table 1 and Table 6.
Examples
>>> import lucid
>>> from lucid.models.vision.mobilenet import mobilenet
>>> model = mobilenet()
>>> x = lucid.randn(1, 3, 224, 224)
>>> out = model(x)
>>> out.last_hidden_state.shape
(1, 1024, 7, 7)