ResUNet 2D

ConvNet Segmentation ConvNet

class lucid.models.ResUNet2d(config: UNetConfig)

ResUNet2d is a residual variant of lucid.models.UNet2d that reuses the same configurable encoder-decoder structure but forces residual blocks throughout the network. It is useful when deeper stage stacks or stronger gradient flow are desirable while keeping the same segmentation-oriented U-Net topology.

For volumetric inputs with shape \((N, C, D, H, W)\), see lucid.models.ResUNet3d.

Class Signature

class ResUNet2d(UNet2d):
    def __init__(self, config: UNetConfig) -> None

Parameters

  • config (UNetConfig): U-Net configuration describing the encoder/decoder stage layout, channel widths, sampling strategy, and segmentation output space. ResUNet2d reuses this config and enforces block=”res”.

Methods

ResUNet2d.forward(x: Tensor) → Tensor | dict[str, Tensor | list[Tensor]]

Examples

Build a ResUNet2d from `from_channels`

import lucid
import lucid.models as models

cfg = models.UNetConfig.from_channels(
    in_channels=3,
    out_channels=2,
    channels=(64, 128, 256, 512),
    num_blocks=(2, 2, 2, 3),
    norm="group",
    act="silu",
    upsample_mode="bilinear",
)
model = models.ResUNet2d(cfg)

x = lucid.random.randn(1, 3, 256, 256)
logits = model(x)
print(logits.shape)  # (1, 2, 256, 256)

Build a Custom ResUNet2d with Stage-Level Configuration

import lucid
import lucid.models as models

cfg = models.UNetConfig(
    in_channels=1,
    out_channels=3,
    encoder_stages=[
        models.UNetStageConfig(channels=32, num_blocks=2),
        models.UNetStageConfig(channels=64, num_blocks=2),
        models.UNetStageConfig(channels=128, num_blocks=3, use_attention=True),
        models.UNetStageConfig(channels=256, num_blocks=3),
    ],
    norm="group",
    act="silu",
    downsample_mode="conv",
    upsample_mode="transpose",
    deep_supervision=True,
)
model = models.ResUNet2d(cfg)

x = lucid.random.randn(2, 1, 256, 256)
out = model(x)
print(out["out"].shape)  # (2, 3, 256, 256)
print(len(out["aux"]))   # 2

Notes