Why does a torchvision pretrained mode perform better than a scratch trained one?

Question

I followed the torchvision tutorial (https://pytorch.org/tutorials/intermediate/torchvision_tutorial.html) and trained a segmentation model on my own dataset of aerial image