mask2former_resnet_18¶

lucid.models.mask2former_resnet_18(num_labels: int, *, pretrained_backbone: bool = False, **config_kwargs: Any) → Mask2Former¶

The mask2former_resnet_18 function builds a Mask2Former model with a ResNet-18 backbone preset.

Total Parameters (num_labels=150): 30,972,567

Function Signature¶

@register_model
def mask2former_resnet_18(
    num_labels: int,
    *,
    pretrained_backbone: bool = False,
    **config_kwargs
) -> Mask2Former

Parameters¶

num_labels (int): Number of segmentation classes.
pretrained_backbone (bool, optional): If True, initializes the backbone from available pretrained weights.
config_kwargs (dict, optional): Additional overrides applied to Mask2FormerConfig.

Returns¶

Mask2Former: Mask2Former model configured with a ResNet-18 backbone.

Example Usage¶

from lucid.models.vision.mask2former import mask2former_resnet_18
import lucid

model = mask2former_resnet_18(num_labels=150)
x = lucid.random.randn(1, 3, 224, 224)
pred = model.predict(x)
print(pred.shape)