Old Content

From Upscale Wiki
This is the latest revision of this page; it has no approved revision.
Jump to navigation Jump to search

Some Content from the Discord that will be preserved here forever.



Texture Maps

Normal Maps
Batch Size
HR Size
Dataset Size
Pretrained Model
Normal Maps Alsa 4 CC BY-NC-SA 4.0 Normal Maps 36K 27 128 ? Custom (Normal Maps) Normal Maps - Skyrim artifacted
Normal Maps - Skyrim artifacted Deorder 4 CC0 Skyrim Normal Maps 145K ? 128 ? Skyrim Normal Maps ?

The first one is based on the second one it was trained, with a higher learning rate and insane n_workers and batch_size values. It is meant to replace the old Normal Map model from Deorder, but without adding BC1 compression to your normal maps.

The second one was trained on Skyrim’s Normal Maps, including compression artifacts, so it will have to be redone.


From LyonHrt: https://drive.google.com/open?id=1RpRueBeJFlkQo-KSCAFILDC8AHsl1xFt a de-dither based model trained at 500000 based on an anime dataset(edited)

From LyonHrt: https://drive.google.com/open?id=1gbJC1sM7BSBhW1e5WMymWlCcPhbGV504 a jpeg artefact removal model attempt at 160000 inters(edited)

From loinne: https://drive.google.com/open?id=1DqdeZRGpkrdi3mxVpEgW9cFLQ4jzGTA6 my attempt on Manga109 by fine-tuning RRDB_ESRGAN_x4.pth - slighly more agressive version of Manga109 by Kingdom


Author: kingdomakrillic

Name: 2x_Faux_1x_Colour_Banding_Removal_model

Link: https://www.mediafire.com/file/436f2cbzio6sgr4/DebandDemonstration.pth/file

Description: "I don't recommend actually using it, as I don't think I have the right combination of settings and dataset. It's just an attempt to show off the versatility of ESRGAN. If you do use it, you should downscale the results, as in addition to being debanded, it'll be upscaled by 2x."


Author: kingdomakrillic

Name: 2x_Faux_1x_DeDither

Scale: faux x1 (x2 nearest neighbor scaling)

Description: Tried to train a model to reduce dithering. It only partially succeeded, possibly because I started out with a smaller dataset and kept adding more and more to it out of desperation

Link: https://www.mediafire.com/file/9xb6td54hrvbck8/De-dither_attempt.zip/file

batch_size: 4

Pretrained_Model_G: None, trained from scratch


Author: loinne

Notice: Not recommended to use (optional)

Scale: 4

Description: Used 9e-2 Gan_weight

Link: https://drive.google.com/open?id=1DS8VrS5PwqyBZ1zzhGNDEE4BSrd_bVw0

Iterations: about 80k

batch_size: 12

Dataset_size: 4700

Pretrained_Model_G: Default PSNR_4X

Other Model Sources

The very first Model Database outside of our Discord from Etokapa. It isn't up to date but includes some models that are not from our community or have been deprecated, so if you are lucky the links in it might still work and you might find a gem. https://docs.google.com/spreadsheets/d/1asDV7NVBlLD25bd4w77mHMxSM7tvexXySafTzSugeHE/edit#gid=0

PPON Models


Pixel Art

Name Author Scale License Purpose Iterations (Phase 1; 2; 3) Batch Size (Phase 1; 2; 3) HR Size (Phase 1; 2; 3) Dataset Size (Phase 1; 2; 3) Dataset (Phase 1; 2; 3) Pretrained Model
Pixie victorca25 4 Pixel Art / some Cartoons 80K(?; ?; ?) 8; 8; 8 192; 192; 192 ?; ?; ? Custom(Drawings); Custom(Drawings); Custom(Drawings) PPON
xBRZ+ victorca25 4 Pixel Art 60K(?; ?; ?) 8; 8; 8 128; 128; 128 ?; ?; ? Custom (xBRZ images); Custom (xBRZ images); Custom (Drawings) Pixie

Pretrained models for different scales

PPON Zheng Hui (惠政) 4



An alternate script with alpha support https://www.dropbox.com/s/uv6hxur4qpr6za5/alpha.py?dl=0

From: Emilien I should really upload my script so you can pass upscale as an argument. It has a few params and I've added the video script from reddit. https://cdn.discordapp.com/attachments/547950274313191425/560137215104253986/test_-_remadewithfolderandvideosupport.py