does anyone has experience running state of the art neural networks in opencv.dnn?

i´m trying to restore some old family photos and plan to use this github projects. i see they have the pretrained models available. can i use the CV2.dnn module to run them inside my own script? and if so do i have to preprocess the images or how do i proceed?

https://github.com/microsoft/Bringing-Old-Photos-Back-to-Lifehttps://github.com/jantic/DeOldifyhttps://github.com/xinntao/Real-ESRGAN