efficientnet_b4

→EfficientNet

efficientnet_b4(pretrained: bool = False, overrides: object = {})

source edit

EfficientNet-B4 feature-extracting backbone (compound coefficient $\phi = 3$ ).

Builds an EfficientNet with width multiplier 1.4 and depth multiplier 1.8. Approximately 19M parameters and 82.9% ImageNet-1k top-1 accuracy at the 380×380 native resolution (Tan & Le, 2019, Table 2).

Model Size

Parameters

pretrainedbool= False

Reserved for future pretrained-weight loading. Currently ignored.

**overridesobject= {}

Keyword overrides forwarded into EfficientNetConfig.

Returns

EfficientNet

Backbone with the B4 configuration applied (or with overrides merged on top of it).

Notes

See Tan & Le, "EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks", ICML 2019 (arXiv:1905.11946), Table 2. Head expansion: round(1280 · 1.4) = 1792 channels.

Examples

>>> import lucid
>>> from lucid.models.vision.efficientnet import efficientnet_b4
>>> model = efficientnet_b4()
>>> x = lucid.randn(1, 3, 380, 380)
>>> out = model(x)
>>> out.last_hidden_state.shape
(1, 1792, 1, 1)

Used by 1

lucid.models.vision.efficientnet

>>> import lucid >>> from lucid.models.vision.efficientnet import efficientnet_b4 >>> model = efficientnet_b4() >>> x = lucid.randn(1, 3, 380, 380) >>> out = model(x) >>> out.last_hidden_state.shape (1, 1792, 1, 1)