Hello,
In one of the lessons for Computer Vision, there is a model that uses the VGG16 and FCN8. What would I need to change if I wanted to use this as a super resolution model? As in, inputs would be low-resolution images and the output would be a high resolution. Output can be the same size or larger. If it’s the same size, I planned on passing the output to the ESRGAN on TF-hub that will increase the size 4x.
Would it be just changing the number of features the FCN8 outputs to 3? I think I tried this a couple weeks ago but wasn’t pleased with the results and said I’d revisit and I think I scrapped that.