EfficientNetEncoder¶

class EfficientNetEncoder(stage_idxs, out_channels, depth=5, channel_multiplier=1.0, depth_multiplier=1.0, drop_rate=0.2, output_stride=32)[source]¶

EfficientNet encoder with configurable scaling parameters.

This class extends EfficientNetBaseEncoder to provide scaling options for depth and channel multipliers, enabling flexible encoder configurations for segmentation and classification tasks.

Features:

  • Supports depth and channel scaling.

  • Provides pretrained weights for multiple variants.

  • Outputs multi-scale feature maps for downstream tasks.

Example

>>> encoder = EfficientNetEncoder(
...     stage_idxs=[2, 3, 5],
...     out_channels=[3, 32, 24, 40, 112, 320],
...     channel_multiplier=1.0,
...     depth_multiplier=1.0,
...     drop_rate=0.2
... )
>>> x = torch.randn(1, 3, 224, 224)
>>> features = encoder(x)
>>> [f.shape for f in features]
... [torch.Size([1, 3, 224, 224]), torch.Size([1, 32, 112, 112]), ...]

Initialize EfficientNetEncoder.

Creates an EfficientNet encoder with configurable scaling parameters for depth and channel multipliers.

Parameters:

  • stage_idxs (list[int]) – Indices of stages for feature extraction.

  • out_channels (list[int]) – Output channels for each depth level.

  • depth (int) – Encoder depth (1-5). Defaults to 5.

  • channel_multiplier (float) – Channel scaling factor. Defaults to 1.0.

  • depth_multiplier (float) – Depth scaling factor. Defaults to 1.0.

  • drop_rate (float) – Dropout rate. Defaults to 0.2.

  • output_stride (int) – Output stride of encoder. Defaults to 32.

Methods

Attributes

training