forked from nagadomi/nunif
-
Notifications
You must be signed in to change notification settings - Fork 0
/
NOTICE
79 lines (65 loc) · 2.63 KB
/
NOTICE
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
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
----
nunif/optim/lion.py
From Google Brain AutoML: https://github.com/google/automl/blob/master/lion/lion_pytorch.py
Google Research / Apache License, Version 2.0
----
nunif/training/weight_decay_config.py
From nanoGPT: https://github.com/karpathy/nanoGPT/blob/master/model.py
Andrej Karpathy / MIT License
----
nunif/waifu2x/docs/images/miku_128.png
Hatsune Miku / Crypton Future Media inc. / CC BY-NC
https://piapro.net/intl/en_for_creators.html
----
nunif/utils/perlin2d.py
The original code is perlin-numpy: https://github.com/pvigier/perlin-numpy
Pierre Vigier / MIT License
Vadim Kantorov ported to pytorch: https://gist.github.com/vadimkantorov/ac1b097753f217c5c11bc2ff396e0a57
----
nunif/training/env.py
The adaptive weight calculation for the two different losses(GAN and others) is from Taming Transformers.
https://github.com/CompVis/taming-transformers
----
nunif/module/_lpips_1.pth
nunif/module/_lpips_2.pth
are trained with the private forked version of LPIPS.
https://github.com/richzhang/PerceptualSimilarity
---
nunif/modules/fourier_unit.py
Modified from LaMa https://github.com/advimman/lama/blob/main/saicinpainting/training/modules/ffc.py
Samsung Research / Apache License Version 2.0
Original implementation Fast Fourier Convolution https://github.com/pkumivision/FFC/blob/main/model_zoo/ffc.py
Lu Chi / Apache License Version 2.0
---
iw3/training/sbs/stereoimage_generation.py
is copied from https://github.com/thygate/stable-diffusion-webui-depthmap-script/blob/2838dc5fa24ac9f9e24b34ff61f2d2c169b634a1/scripts/stereoimage_generation.py
Bob Thiry / MIT License
And contributors https://github.com/thygate/stable-diffusion-webui-depthmap-script/graphs/contributors
---
iw3 uses forked version of ZoeDepth and MiDaS as its core logic
https://github.com/isl-org/MiDaS
https://github.com/isl-org/ZoeDepth
Intelligent Systems Lab Org / MIT License
---
iw3 uses forked version of Depth-Anything
https://github.com/LiheYoung/Depth-Anything
Lihe Yang / Apache License Version 2.0
---
iw3 can use U-2-Net's pretrained model via rembg module.
https://github.com/xuebinqin/U-2-Net
xuebinqin / Apache License Version 2.0
---
iw3 can use DIS's pretrained model via rembg module.
https://github.com/xuebinqin/DIS
xuebinqin / Apache License Version 2.0
---
iw3 can anime-segmentation's pretrained model via rembg module.
https://github.com/SkyTNT/anime-segmentation
SkyTNT / Apache License Version 2.0
---
nunif/modules/lpips.py
`normalize_tensor` fix is adapted from NormFixLPIPS of NeuralCompression
https://github.com/facebookresearch/NeuralCompression
Meta Platforms, Inc. and affiliates / MIT License
----
Other libraries used are listed in `requirements.txt`.