Can I use edge detection filter (exp: Sobel) to implement it in an unsupervised CNN model ?

I want to use Sobel filter with a CNN model to make an unsupervised learning model to extract edges from the images, I read an article that made a model using this architecture: input layer - 3 hidden layers (Sobel and derivatives of gaussian are used) - output layer.

I wanted to ask if such a model could be made and how since I made a lot of researches and couldn't find something similar.