Usage of HRNet in convolutional autoencoders for image generation

[sparshgarg23]

Hi,
I was looking into resnet based convolution auto encoders which are used for image reconstruction and generation.
Since HRNet is able to maintain the high resolution as well as spatial precision,would it be possible to use HRNet for image generation purposes.
If it can be used for such purposes,then what changes would have to be made in the architecture so that it can be used for that purpose