Hey there,
I work on some AI/ML image generation software with my company and was wondering if this idea made any sense, or if it is already implemented.
Would it be beneficial to create actual data sets for DLSS targeted for specific games.
The idea is this: Create/Use an automation tool that takes screenshots at highest resolution and highest quality for a specific game and train DLSS with this data. When DLSS is used for that game it utilizes the additional data set which “knows” what the game is “supposed” to look like.
Not sure what data DLSS is already trained on, and if this is something that already happens.
What do you all think?