efficientnet_b4_cls

→EfficientNetForImageClassification

efficientnet_b4_cls(pretrained: bool | str = False, weights: EfficientNetB4Weights | None = None, overrides: object = {})

source edit

EfficientNet-B4 image classifier (backbone + GAP + linear head).

Builds an EfficientNetForImageClassification with the B4 compound-scaling configuration (width × 1.4, depth × 1.8) followed by dropout ( $p = 0.4$ ) and a linear classifier. Approximately 19M parameters and 82.9% ImageNet-1k top-1 accuracy at 380×380 (Tan & Le, 2019, Table 2).

Model Size

Parameters

pretrainedbool or str= False

Pretrained-weight selector. False → random init; True → the DEFAULT tag (EfficientNetB4Weights.IMAGENET1K_V1); a tag string → that specific checkpoint. Mutually exclusive with weights.

weights(EfficientNetB4Weights, optional, keyword - only)= None

Explicit weights enum member. Takes precedence over pretrained.

**overridesobject= {}

Keyword overrides forwarded into EfficientNetConfig.

Returns

EfficientNetForImageClassification

Classifier with the B4 configuration applied (or with overrides merged on top of it), optionally initialised from pretrained weights.

Notes

See Tan & Le, "EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks", ICML 2019 (arXiv:1905.11946), Table 2. Pretrained weights are converted from the reference-framework EfficientNet_B4_Weights and hosted under lucid-dl/efficientnet-b4.

Examples

>>> import lucid
>>> from lucid.models.vision.efficientnet import efficientnet_b4_cls
>>> model = efficientnet_b4_cls(pretrained=True)
>>> x = lucid.randn(1, 3, 380, 380)
>>> out = model(x)
>>> out.logits.shape
(1, 1000)

Used by 1

lucid.models.vision.efficientnet

>>> import lucid >>> from lucid.models.vision.efficientnet import efficientnet_b4_cls >>> model = efficientnet_b4_cls(pretrained=True) >>> x = lucid.randn(1, 3, 380, 380) >>> out = model(x) >>> out.logits.shape (1, 1000)