Pretrained weights for lucid.models.se_resnet_50_cls.
Sourced from timm's seresnet50.ra2_in1k RandAugment recipe — a
cleanly SE-augmented ResNet whose state-dict naming matches Lucid
one-for-one (identity key map).
Attributes
RA2_IN1KWeightEntrytimm RandAugment-v2 ImageNet-1k checkpoint (top-1 78.498%),
sourced from
timm/seresnet50.ra2_in1k.DEFAULTWeightEntryAlias for
RA2_IN1K.Notes
Reference: Hu, Shen, Sun, "Squeeze-and-Excitation Networks", CVPR 2018 (arXiv:1709.01507).
Examples
>>> from lucid.models import se_resnet_50_cls
>>> model = se_resnet_50_cls(pretrained=True).eval()