You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
1. Filters trained on 675 images of pexels with area & bicubic & lancoz & nearest & blur degradation
2
-
2. Angle=24 Strength=3 Coherence=3 setting for both filters
1
+
1. Filters trained on 675 images of pexels with area & bicubic & lancoz & nearest & blur degradation.
2
+
2. Angle=24 Strength=3 Coherence=3 setting for both filters.
3
3
3. Cheapupscale: using output with bilinear upscaling for 1st pass, No cheapupscale of 2nd pass training.
4
-
4. 10bit 1pass filter trained with multi-lr: bicubic+area+nearest+lanczos as LR and HRSharpen50 as HR
5
-
10bit 2pass filter trained with HR after single augment:blur as cheapupscale images and sharpened HR images as HR
6
-
8bit 1pass filter trained with multi-lr: bicubic+area+nearest+lanczos as LR and HRSharpen50 as HR
7
-
8bit 2pass filter is the same with the filter4-2pass filter of the past version.
4
+
4. 10bit 1pass filter uses the multi-lr: bicubic+area+nearest+lanczos as LR and higher sharpened HR as reference (s=1.5).
5
+
10bit 2pass filter uses the HR after single augment:blur as cheapupscale images and higher sharpened HR as reference (s=1.5).
6
+
8bit 1pass filter uses the multi-lr: bicubic+area+nearest+lanczos as LR and higher sharpened HR as reference (s=1.5) as reference.
7
+
8bit 2pass filter uses the output images of the filter trained using bilinear upscaling & HR as cheapupscale images and uses higher sharpened HR as reference (s=1.5).
0 commit comments