maskformer_resnet_18¶

lucid.models.maskformer_resnet_18(num_labels: int, *, pretrained_backbone: bool = False, **config_kwargs: Any) → MaskFormer¶

The maskformer_resnet_18 function builds a MaskFormer model with a ResNet-18 backbone preset.

Total Parameters (num_labels=150): 24,700,119

Function Signature¶

@register_model
def maskformer_resnet_18(
    num_labels: int,
    *,
    pretrained_backbone: bool = False,
    **config_kwargs
) -> MaskFormer

Parameters¶

num_labels (int): Number of semantic classes for segmentation.
pretrained_backbone (bool, optional): If True, initializes the ResNet-18 backbone with pretrained classification weights.
config_kwargs (dict, optional): Additional overrides applied to MaskFormerConfig.

Returns¶

MaskFormer: MaskFormer model configured with a ResNet-18 backbone.

Example Usage¶

from lucid.models.vision.maskformer import maskformer_resnet_18
import lucid

model = maskformer_resnet_18(num_labels=150)
x = lucid.random.randn(1, 3, 512, 512)
pred = model.predict(x)
print(pred.shape)