File size: 1,559 Bytes
d4b55f0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Name: 4xRealWebPhoto_RGT
License: CC BY 4.0
[Google Drive Folder](https://drive.google.com/drive/folders/1stOfI5a2W8KTI-FAbhFhmkYo2JEtjoGi?usp=drive_link)
Author: Philip Hofmann
Network: RGT
Scale: 4
Release Date: 10.03.2024
Purpose: 4x real web photo upscaler, meant for upscaling photos downloaded from the web
Iterations: 260'000
epoch: 71
batch_size: 12
HR_size: 128
Dataset: 4xRealWebPhoto
Number of train images: 6118
OTF Training: No
Pretrained_Model_G: RGT_x4

Description:
4x real web photo upscaler, meant for upscaling photos downloaded from the web. Trained on my v1 of my 4xRealWebPhoto dataset, it should be able to handle noise, jpg and webp (re)compression, (re)scaling, and a bit of lens blur. I tried to simulate the use case of someone uploading a photo (maybe noisy, maybe blurry, maybe noisy&blurry) to the web (for example social media) where the provider would downscale&compress the image, and then another user downloading and re-uploading it on the web (recompression & rescale).

Workflow for dataset degradation can be found [in this pdf](https://drive.google.com/file/d/1JmB4yxGtZQAtFwKn3DmKAPzfjC2SZNQ1/view?usp=drive_link) and models can be found [in this google drive folder](https://drive.google.com/drive/folders/1stOfI5a2W8KTI-FAbhFhmkYo2JEtjoGi?usp=drive_link). v0 was basically trained on a degraded RSBlur dataset with strong motion blur, v1 (this one) was on nomos8k_sfw, and v2 was on nomos8k with multiscale and variants, also different degradation values.  

12 Examples:   
[Slowpics](https://slow.pics/s/HxXtE1eZ)