class

EfficientNetB3Weights

extendsWeightsEnum

EfficientNetB3Weights()

Pretrained weights for lucid.models.efficientnet_b3_cls.

compound-scaled B3 at 300 (12.2 M params, 1.83 GFLOPs, top-1 82.01%).

IMAGENET1K_V1WeightEntry

ImageNet-1k V1 checkpoint (top-1 82.008% / top-5 96.054%), sourced from torchvision/EfficientNet_B3_Weights.IMAGENET1K_V1.

DEFAULTWeightEntry

Alias for IMAGENET1K_V1.

Reference: Tan, Le, "EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks", ICML 2019 (arXiv:1905.11946).

>>> from lucid.models import efficientnet_b3_cls
>>> model = efficientnet_b3_cls(pretrained=True).eval()