EfficientNet-B4

Tan & Le, 2019 — EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks (arXiv:1905.11946)

Lucid port of torchvision/EfficientNet_B4_Weights.IMAGENET1K_V1, converted to Lucid-native safetensors.

Available weights

Tag	acc@1	acc@5	Params	GFLOPs	Size	Source
IMAGENET1K_V1 (default)	83.384	96.594	19.3M	4.394	74.33 MB	torchvision

Usage

import lucid.models as models
from lucid.models.weights import EfficientNetB4Weights

# default tag
model = models.efficientnet_b4_cls(pretrained=True)

# explicit tag (enum or string)
model = models.efficientnet_b4_cls(weights=EfficientNetB4Weights.IMAGENET1K_V1)
model = models.efficientnet_b4_cls(pretrained="IMAGENET1K_V1")

# preprocessing travels with the weights
weights = EfficientNetB4Weights.IMAGENET1K_V1
preprocess = weights.transforms()
logits = model(preprocess(image)[None]).logits

Conversion

Converted from torchvision/EfficientNet_B4_Weights.IMAGENET1K_V1 via python -m tools.convert_weights efficientnet_b4 --tag IMAGENET1K_V1. Key mapping + numerical parity verified against the source.

License

apache-2.0 — inherited from the original weights.

Citation

@inproceedings{tan2019efficientnet,
  title={EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks},
  author={Tan, Mingxing and Le, Quoc},
  booktitle={ICML}, year={2019}
}