ResNet

ResNet模型基于Deep Residual Learning for Image Recognition论文。

注意

TorchVision的瓶颈将下采样的步幅放置在第二个3x3卷积中，而原始论文将其放置在第一个1x1卷积中。这种变体提高了准确性，被称为ResNet V1.5。

模型构建器

以下模型构建器可用于实例化ResNet模型，无论是否使用预训练权重。所有模型构建器内部都依赖于torchvision.models.resnet.ResNet基类。有关此类的更多详细信息，请参阅源代码。

resnet18(*[, weights, progress])	ResNet-18 来自 Deep Residual Learning for Image Recognition.
resnet34(*[, weights, progress])	ResNet-34 来自 Deep Residual Learning for Image Recognition.
resnet50(*[, weights, progress])	ResNet-50 来自 Deep Residual Learning for Image Recognition.
resnet101(*[, weights, progress])	ResNet-101 来自 Deep Residual Learning for Image Recognition.
resnet152(*[, weights, progress])	ResNet-152 来自 Deep Residual Learning for Image Recognition.