MaxViT-Base backbone (Tu et al., 2022).
Builds the canonical MaxViT-Base configuration:
depths=(2, 6, 14, 2), dims=(96, 192, 384, 768).
118.1M parameters headless (119.5M with the classifier
head; paper Table 1 lists 119M).
Model Size
Parameters
pretrainedbool= FalseNo checkpoint is published for
maxvit_base, so True raises
NotImplementedError rather than return randomly initialised weights.
maxvit_base_cls is the checkpointed variant.**overridesobject= {}Keyword overrides on top of the canonical MaxViT-Base config.
Returns
MaxViTA MaxViT backbone returning a flat
feature.
Notes
MaxViT-Base reaches 84.9% top-1 on ImageNet-1k at 224x224 (Tu et al., 2022, Table 6).
Examples
>>> import lucid
>>> from lucid.models.vision.maxvit import maxvit_base
>>> model = maxvit_base()
>>> x = lucid.randn(1, 3, 224, 224)
>>> model.forward_features(x).shape
(1, 768)