Hello
Thanks for reading my blog and glad you liked it!

So, UNet with EfficientNetB0 as Encoder means — In unet model, we have 2 parts, encoder and decoder. Encoder extracts all the features from the images by performing convolution operation and then decoder uses deconvolution to make the size of these extracted feature map equal to the size of input image.

Here, encoder can be any combination of Conv and Pooling layers. But, here I have used EfficientNetB0 architecture as encoder. This means, instead of using random set of Conv and Pooling, I am following Conv and Poolings of EfficientNetB0 architecture. Similarly, you can use ResNet or VGG as encoder here.