We should talk about OCR! left: original, right: Rhea XL
Interesting is, Rhea V1 does it much better
Rhea XL has positive aspects, often it really keeps more details and has the potential to replace Iris and Proteus completly in the future. But in my opinion it’s in beta stage and there is a lot to do. Sometimes the model has problems with background and foreground detection and face recovery does not exist…or does it a little bit, but can’t reach Iris.
Thanks for sharing the results here, did you run this on auto? Will want to share the details with the research team.
Hi, thanks for you’re response! Yes I did it in “Auto Mode” without additional settings. 4x upscale of 720 x 576 progressive content (source 4:3 flaged “display content”, setting “square pixels” in Topaz). I can generate some more samples if it helps.
Here additional samples: 4x Rhea XL auto upscale without any user settings. in the second part of the video I have overlayed some fonts for testing what Rhea XL does with this.
and here is the used video ziped
fonts_Rhea.zip (14.1 MB)
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.
