Skip to content

Releases: HelpSeeker/FSRCNN-TensorFlow

FSRCNNX-Distort Final Version

15 Feb 07:40
24ba21f
Compare
Choose a tag to compare

For general infos see igv's release page.

This is the final release.

I haven't touched this project in ages, but there are two additional shaders that I've never come around to share. They are x1 versions of the previous two shaders and will get applied no matter the scaling factor. Think of them more like filters.

All these networks were trained with increased distortion and as a result remove more compression artifacts than igv's. This also means they aren't as sharp and remove details.

These shaders are intended for low quality content. And I mean LOW quality. The goal of these shaders is to make ultra compressed sub-SD videos a bit more bearable to watch.
Please keep this in mind, when you see somebody compare the results of this project to igv's or Anime4K's.

The anime version offers stronger compression artifact removal and line emphasis and kills even more details in the process. It is definitely not suited for real life content. Even for anime it is hit or miss. I recommend to stick with the regular versions.

FSRCNNX-Distort

03 Oct 19:20
ef687e6
Compare
Choose a tag to compare

For general infos see igv's release page.

These networks were trained with increased distortion and as a result remove quite a bit more compression artifacts than igv's. This also means that they aren't as sharp and will remove small details. Don't use these shaders for high quality footage.

The anime version offers even stronger compression artifact removal. It is not suited for real life content, unless you feel like your videos lack a certain pastel painting look.