How to extract feature from resnet autoencoder?

I want to use resnet autoencoder for construction an image. ie I have an image and I want to reconstruction that by using resnet autoencoder and then I need to have features that extract from encoder that used in autoencoder.