Phips commited on
Commit
5c20f85
·
verified ·
1 Parent(s): 8398fc9

Upload 3 files

Browse files
weights/2xNomosUni_compact_multijpg_ldl.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3127e5fe3085b13765314eb643a0654afc1a596b6d948af2be1e790299ce3365
3
+ size 2419161
weights/2xNomosUni_compact_otf_medium.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d9f767194d5163e5b1604ccd87f24f12f23562ed66245bba058eea1f4686ca5a
3
+ size 2419158
weights/info.txt ADDED
@@ -0,0 +1,57 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ See the information about the two models below:
2
+ Author: Philip Hofmann
3
+ License: CC BY 4.0
4
+ Release Date: 11.01.2024 (dd/mm/yy)
5
+ Trained with musl's [neosr](https://github.com/muslll/neosr), kim's [Datasetdestroyer](https://github.com/Kim2091/helpful-scripts/tree/main/Dataset%20Destroyer), and the Real-ESRGAN pipeline.
6
+
7
+ ---
8
+
9
+ Name: 2xNomosUni_compact_multijpg_ldl
10
+ Author: Philip Hofmann
11
+ Release Date: 11.01.2024 (dd/mm/yy)
12
+ License: CC BY 4.0
13
+ Network: SRVGGNetCompact
14
+ Scale: 2
15
+ Purpose: 2x fast universal DoF preserving upscaler
16
+ Iterations: 218'000
17
+ epoch: 168
18
+ batch_size: 12
19
+ HR_size: 128
20
+ Dataset: nomosuni
21
+ Number of train images: 2989
22
+ OTF Training: No
23
+ Pretrained_Model_G: 2x-Compact-Pretrain
24
+
25
+ Description:
26
+ 2x compact fast universal DoF preserving upscaler, pair trained with jpg degradation (down to 40) and multiscale (down_up, bicubic, bilinear, box, nearest, lanczos).
27
+
28
+ Examples:
29
+ [ani](https://slow.pics/c/Zd7UCmLP)
30
+ [face](https://slow.pics/c/wKiu7Orf)
31
+ [dofwheat](https://imgsli.com/MjMyNTU4)
32
+
33
+ ---
34
+
35
+ Name: 2xNomosUni_compact_otf_medium
36
+ Author: Philip Hofmann
37
+ Release Date: 11.01.2024 (dd/mm/yy)
38
+ License: CC BY 4.0
39
+ Network: SRVGGNetCompact
40
+ Scale: 2
41
+ Purpose: 2x fast universal upscaler with medium degradation handling (jpg compression, noise, blur)
42
+ Iterations: 276'000
43
+ epoch: 218
44
+ batch_size: 12
45
+ HR_size: 128
46
+ Dataset: nomosuni
47
+ Number of train images: 2989
48
+ OTF Training: Yes
49
+ Pretrained_Model_G: 2xNomosUni_compact_otf_strong
50
+
51
+ Description:
52
+ 2x compact fast universal upscaler with medium degradation handling using the Real-ESRGAN training pipeline, based off 2xNomosUni_compact_otf_strong. Handles jpg compression, some noise, and some blur (so dejpgs, denoises and deblurs).
53
+
54
+ Examples:
55
+ [RealPhoto](https://imgsli.com/MjMyNTUy)
56
+ [Noisy](https://imgsli.com/MjMyNTUz)
57
+