Adding dilated convolutions to VGG16

vision
1

I want to add dilated convolutions to the 5th convolutional block(all three layers) of VGG16 and finetune the pretrained weights. I wanted to know what the best method is to go about it.

2

Would you like to add these conv layers between the other layers or would you like to change the dilation parameter of the conv layers in the vgg model?