PVT v2-B4 backbone (Wang et al., 2022).
Builds the canonical PVT v2-B4 configuration:
embed_dims=(64, 128, 320, 512), depths=(3, 8, 27, 3).
Approximately 62.6M parameters.
Model Size
Parameters
pretrainedbool= FalseNo checkpoint is published for
pvt_v2_b4, so True raises
NotImplementedError rather than return randomly initialised weights.
pvt_v2_b4_cls is the checkpointed variant.**overridesobject= {}Keyword overrides on top of the canonical PVT v2-B4 config.
Returns
PVTA PVT backbone returning a flat
feature.
Notes
PVT v2-B4 reaches 83.6% top-1 on ImageNet-1k at 224x224 (Wang et al., 2022, Table 1).
Examples
>>> import lucid
>>> from lucid.models.vision.pvt import pvt_v2_b4
>>> model = pvt_v2_b4()
>>> x = lucid.randn(1, 3, 224, 224)
>>> model.forward_features(x).shape
(1, 512)