Pretrained weights for lucid.models.vgg_16_bn_cls.
VGG-D with BatchNorm (138.4 M params, 15.5 GFLOPs, top-1 73.36%).
Attributes
IMAGENET1K_V1WeightEntryImageNet-1k V1 checkpoint (top-1 73.360% / top-5 91.516%),
sourced from
torchvision/VGG16_BN_Weights.IMAGENET1K_V1.DEFAULTWeightEntryAlias for
IMAGENET1K_V1.Notes
Reference: Simonyan, Zisserman, "Very Deep Convolutional Networks for Large-Scale Image Recognition", ICLR 2015 (arXiv:1409.1556). BatchNorm added in the torchvision reimplementation (not in the original paper).
Examples
>>> from lucid.models import vgg_16_bn_cls
>>> model = vgg_16_bn_cls(pretrained=True).eval()