How can i modify the feature map size parameter in ResNET50?

vision
1

I want to change the output feature map size of layer4 of ResNet50,

    self.layer1 = self._make_layer(block, 64, layers[0])
        self.layer2 = self._make_layer(block, 128, layers[1], stride=2,
                                       dilate=replace_stride_with_dilation[0])
        self.layer3 = self._make_layer(block, 256, layers[2], stride=2,
                                       dilate=replace_stride_with_dilation[1])
        self.layer4 = self._make_layer(block, 512, layers[3], stride=2,
                                       dilate=replace_stride_with_dilation[2])
        self.avgpool = nn.AdaptiveAvgPool2d((1, 1))

here, i want to change

self.layer4 = self._make_layer(block, 512, layers[3], stride=2,
                                       dilate=replace_stride_with_dilation[2])

to

self.layer4 = self._make_layer(block, 256, layers[3], stride=2,
                                       dilate=replace_stride_with_dilation[2])

but this is built-in Model so I have no idea to change that. How can i change this ?

2

You can specify any model inside of the class as the parent. And then modify whichever layer you want in the init. See here: