se_resnet_34¶
Overview¶
The se_resnet_34 function constructs an SE-ResNet-34 model, a residual network with SE blocks for channel-wise feature recalibration, designed for moderately complex image classification tasks.
Total Parameters: 21,958,868
Function Signature¶
@register_model
def se_resnet_34(num_classes: int = 1000, **kwargs) -> SENet:
Parameters¶
num_classes (int, optional): Number of output classes for the classification task. Default is 1000.
kwargs: Additional keyword arguments to customize the model.
Returns¶
SENet: An instance of the SE-ResNet-34 model.
Examples¶
Creating an SE-ResNet-34 model for 1000 classes:
model = se_resnet_34(num_classes=1000)
print(model)
Note
SE-ResNet-34 uses a configuration of [3, 4, 6, 3] for its layers.
Incorporates _SEResNetModule for SE operations.
Initializes weights internally unless specified otherwise through kwargs.