Create and train a UNet-ResNet deep convolutional neural network for image segmentation..

You will be given access to the QuakeCity dataset consisting of computer generated synthetic images of earthquake damaged buildings developed at SAIL-UH https://sail.cive.uh.edu/quakecity/. You will create and train a UNet-ResNet deep convolutional neural network for image segmentation. Each pixel of every image has 8 possible labels, each label is a component of a structure such as the windows, balcony, and columns, etc. The data set consists of 2130 training images, 914 validation images, and 1004 images for testing. The testing images will be used to calculate your mean intersection over union.