Project Collection
Self-Trained Upscaling Models
Overview
After I had created and worked on my interactive upscaling models comparison website, I wanted to learn how to create such models myself. These are some models that I have trained and published on discord so far:
4xLSDIRCompact (v1-3)
2xHFA2kCompact
2xParimgCompact
4xHFA2x
Approach
I went on the Enhance Everything discord server and started reading the discussions there. I watched youtube videos that explained the whole GAN concepts more. I read github repo readmes. I found a how to and tried around installing dependencies and getting things to work on my Ubuntu system until training worked. I found the LSDIR dataset and started using that. Getting feedback and having discussions on discord I started learning more and improving my configs. Had used (learned to use) Google Colab, Kaggle and Runpod to extend local resources (especially using GPU's with more VRAM than mine). Published trained models on discord.