Skip to content

Commit b65c118

Browse files
Kim2091Kim2091
authored
add animesharpv2 + pbrify (#462)
1 parent 84f0c3c commit b65c118

6 files changed

+311
-1
lines changed

‎data/collections.json

Lines changed: 13 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,15 @@
11
{
2+
"c-animesharpv2": {
3+
"name": "AnimeSharpV2",
4+
"author": "kim2091",
5+
"description": "",
6+
"models": [
7+
"2x-AnimeSharpV2-RPLKSR-Sharp",
8+
"2x-AnimeSharpV2-RPLKSR-Soft",
9+
"2x-AnimeSharpV2-MoSR-Sharp",
10+
"2x-AnimeSharpV2-MoSR-Soft"
11+
]
12+
},
213
"c-normal-map-upscaling": {
314
"name": "Normal Map Upscaling",
415
"description": "This collection contain my RG0 normal map upscaling models.\n\nAll models here are for upscaling *tangent-space* normal maps in RG0 format. RG0 means that the B channel is set to 0. These models will work not correctly if you give them images with non-zero B channel, so you either have to zero the B channel manually or use tool like chaiNNer to do it.\n\n## DDS Compression\n\nI made 3 versions: \n- Normal RG0 is for uncompressed normal map textures. Since it hasn't been trained on compression artifacts, it's highly sensitive to quantization artifacts and noise.\n- Normal RG0 BC1 is for BC1-compressed DDS normal map textures.\n- Normal RG0 BC7 is for BC7-compressed DDS normal map textures. This model sometimes produces images that aren't as sharp. In those cases, you can try the BC1 version to see whether it gives better results.",
@@ -18,7 +29,8 @@
1829
"1x-PBRify-RoughnessV2",
1930
"4x-PBRify-UpscalerSPANV4",
2031
"4x-PBRify-UpscalerSIR-M-V2",
21-
"4x-PBRify-UpscalerDAT2-V1"
32+
"4x-PBRify-UpscalerDAT2-V1",
33+
"4x-PBRify-RPLKSRd-V3"
2234
],
2335
"author": "kim2091"
2436
}

‎data/models/2x-AnimeSharpV2-MoSR-Sharp.json

Lines changed: 60 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,60 @@
1+
{
2+
"name": "2x-AnimeSharpV2_MoSR_Sharp",
3+
"author": "kim2091",
4+
"license": "CC-BY-NC-SA-4.0",
5+
"tags": [
6+
"anime",
7+
"compression-removal",
8+
"restoration"
9+
],
10+
"description": "GitHub Link: https://github.com/Kim2091/Kim2091-Models/releases/tag/2x-AnimeSharpV2_Set\n\nThis is my first anime model in years. Hopefully you guys can find a good use-case for it. Included are 4 models:\n\n1. RealPLKSR (Higher quality, slower)\n2. MoSR (Lower quality, faster)\n\nThere are Sharp and Soft versions of both\nWhen to use each:\n- __Sharp:__ For heavily degraded sources. Sharp models have issues depth of field but are best at removing artifacts \n- __Soft:__ For cleaner sources. Soft models preserve depth of field but may not remove other artifacts as well\n\n__Notes:__\n- MoSR doesn't work in chaiNNer currently\n- To use MoSR:\n 1. Use the ONNX version in tools like [VideoJaNai](<https://github.com/the-database/VideoJaNai>)\n 2. Update spandrel in the latest version of ComfyUI\n\nThe ONNX version may produce slightly different results than the .pth version. If you have issues, try the .pth model.",
11+
"date": "2024-10-05",
12+
"architecture": "mosr",
13+
"size": null,
14+
"scale": 2,
15+
"inputChannels": 3,
16+
"outputChannels": 3,
17+
"resources": [
18+
{
19+
"platform": "pytorch",
20+
"type": "pth",
21+
"size": 17324914,
22+
"sha256": "5a69d1c681aef2251802e69131631868b451c3874e0afb46f55bd6cc820e6400",
23+
"urls": [
24+
"https://github.com/Kim2091/Kim2091-Models/releases/download/2x-AnimeSharpV2_Set/2x-AnimeSharpV2_MoSR_Sharp.pth"
25+
]
26+
},
27+
{
28+
"platform": "onnx",
29+
"type": "onnx",
30+
"size": 8844378,
31+
"sha256": "027cd4d14b5c9ef860cb74d7c17b45f6fa84cfe916e75378f6ad78d554afb6d4",
32+
"urls": [
33+
"https://github.com/Kim2091/Kim2091-Models/releases/download/2x-AnimeSharpV2_Set/2x-AnimeSharpV2_MoSR_Sharp_fp16.onnx"
34+
]
35+
}
36+
],
37+
"trainingIterations": 150000,
38+
"trainingBatchSize": 10,
39+
"trainingHRSize": 256,
40+
"trainingOTF": false,
41+
"dataset": "HFA2k Modified",
42+
"datasetSize": 3000,
43+
"images": [
44+
{
45+
"type": "paired",
46+
"LR": "https://i.slow.pics/7JZs9Otd.webp",
47+
"SR": "https://i.slow.pics/Sz54sRjY.webp"
48+
},
49+
{
50+
"type": "paired",
51+
"LR": "https://i.slow.pics/mJZnrE4U.webp",
52+
"SR": "https://i.slow.pics/kfXY7KhO.webp"
53+
},
54+
{
55+
"type": "paired",
56+
"LR": "https://i.slow.pics/4RUqHkln.webp",
57+
"SR": "https://i.slow.pics/8A1U5Fwf.webp"
58+
}
59+
]
60+
}

‎data/models/2x-AnimeSharpV2-MoSR-Soft.json

Lines changed: 60 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,60 @@
1+
{
2+
"name": "2x-AnimeSharpV2_MoSR_Soft",
3+
"author": "kim2091",
4+
"license": "CC-BY-NC-SA-4.0",
5+
"tags": [
6+
"anime",
7+
"compression-removal",
8+
"restoration"
9+
],
10+
"description": "GitHub Link: https://github.com/Kim2091/Kim2091-Models/releases/tag/2x-AnimeSharpV2_Set\n\nThis is my first anime model in years. Hopefully you guys can find a good use-case for it. Included are 4 models:\n\n1. RealPLKSR (Higher quality, slower)\n2. MoSR (Lower quality, faster)\n\nThere are Sharp and Soft versions of both\nWhen to use each:\n- __Sharp:__ For heavily degraded sources. Sharp models have issues depth of field but are best at removing artifacts \n- __Soft:__ For cleaner sources. Soft models preserve depth of field but may not remove other artifacts as well\n\n__Notes:__\n- MoSR doesn't work in chaiNNer currently\n- To use MoSR:\n 1. Use the ONNX version in tools like [VideoJaNai](<https://github.com/the-database/VideoJaNai>)\n 2. Update spandrel in the latest version of ComfyUI\n\nThe ONNX version may produce slightly different results than the .pth version. If you have issues, try the .pth model.",
11+
"date": "2024-10-05",
12+
"architecture": "mosr",
13+
"size": null,
14+
"scale": 2,
15+
"inputChannels": 3,
16+
"outputChannels": 3,
17+
"resources": [
18+
{
19+
"platform": "pytorch",
20+
"type": "pth",
21+
"size": 17324914,
22+
"sha256": "141bd9b90c323f84cbeb17b4238f3b27df29fb43aeb09916aa6791d00e9352e4",
23+
"urls": [
24+
"https://github.com/Kim2091/Kim2091-Models/releases/download/2x-AnimeSharpV2_Set/2x-AnimeSharpV2_MoSR_Soft.pth"
25+
]
26+
},
27+
{
28+
"platform": "onnx",
29+
"type": "onnx",
30+
"size": 8844378,
31+
"sha256": "e29db0e4b50e0a09b929ad9014ee455802d9f493a82d3d542e6accfccff42743",
32+
"urls": [
33+
"https://github.com/Kim2091/Kim2091-Models/releases/download/2x-AnimeSharpV2_Set/2x-AnimeSharpV2_MoSR_Soft_fp16.onnx"
34+
]
35+
}
36+
],
37+
"trainingIterations": 150000,
38+
"trainingBatchSize": 10,
39+
"trainingHRSize": 256,
40+
"trainingOTF": false,
41+
"dataset": "HFA2k Modified",
42+
"datasetSize": 3000,
43+
"images": [
44+
{
45+
"type": "paired",
46+
"LR": "https://i.slow.pics/7JZs9Otd.webp",
47+
"SR": "https://i.slow.pics/UHvWgH7C.webp"
48+
},
49+
{
50+
"type": "paired",
51+
"LR": "https://i.slow.pics/mJZnrE4U.webp",
52+
"SR": "https://i.slow.pics/x0uFFn4T.webp"
53+
},
54+
{
55+
"type": "paired",
56+
"LR": "https://i.slow.pics/4RUqHkln.webp",
57+
"SR": "https://i.slow.pics/4jyrfzDy.webp"
58+
}
59+
]
60+
}

‎data/models/2x-AnimeSharpV2-RPLKSR-Sharp.json

Lines changed: 60 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,60 @@
1+
{
2+
"name": "2x-AnimeSharpV2_RPLKSR_Sharp",
3+
"author": "kim2091",
4+
"license": "CC-BY-NC-SA-4.0",
5+
"tags": [
6+
"anime",
7+
"compression-removal",
8+
"restoration"
9+
],
10+
"description": "GitHub Link: https://github.com/Kim2091/Kim2091-Models/releases/tag/2x-AnimeSharpV2_Set\n\nThis is my first anime model in years. Hopefully you guys can find a good use-case for it. Included are 4 models:\n\n1. RealPLKSR (Higher quality, slower)\n2. MoSR (Lower quality, faster)\n\nThere are Sharp and Soft versions of both\nWhen to use each:\n- __Sharp:__ For heavily degraded sources. Sharp models have issues depth of field but are best at removing artifacts \n- __Soft:__ For cleaner sources. Soft models preserve depth of field but may not remove other artifacts as well\n\n__Notes:__\n- MoSR doesn't work in chaiNNer currently\n- To use MoSR:\n 1. Use the ONNX version in tools like [VideoJaNai](<https://github.com/the-database/VideoJaNai>)\n 2. Update spandrel in the latest version of ComfyUI\n\nThe ONNX version may produce slightly different results than the .pth version. If you have issues, try the .pth model.",
11+
"date": "2024-10-05",
12+
"architecture": "realplksr",
13+
"size": null,
14+
"scale": 2,
15+
"inputChannels": 3,
16+
"outputChannels": 3,
17+
"resources": [
18+
{
19+
"platform": "pytorch",
20+
"type": "pth",
21+
"size": 29581322,
22+
"sha256": "ff5230dec962235e2ffbc542c232ba438537aefe6cb2db8d072c7bdf1247fbc9",
23+
"urls": [
24+
"https://github.com/Kim2091/Kim2091-Models/releases/download/2x-AnimeSharpV2_Set/2x-AnimeSharpV2_RPLKSR_Sharp.pth"
25+
]
26+
},
27+
{
28+
"platform": "onnx",
29+
"type": "onnx",
30+
"size": 15324988,
31+
"sha256": "b8704239f9cbacec75f6a078257bb2ee7a9a0ee7917c7029a331d1fbf4d59054",
32+
"urls": [
33+
"https://github.com/Kim2091/Kim2091-Models/releases/download/2x-AnimeSharpV2_Set/2x-AnimeSharpV2_RPLKSR_Sharp_fp16.onnx"
34+
]
35+
}
36+
],
37+
"trainingIterations": 150000,
38+
"trainingBatchSize": 10,
39+
"trainingHRSize": 256,
40+
"trainingOTF": false,
41+
"dataset": "HFA2k Modified",
42+
"datasetSize": 3000,
43+
"images": [
44+
{
45+
"type": "paired",
46+
"LR": "https://i.slow.pics/7JZs9Otd.webp",
47+
"SR": "https://i.slow.pics/6Jxhk9W7.webp"
48+
},
49+
{
50+
"type": "paired",
51+
"LR": "https://i.slow.pics/mJZnrE4U.webp",
52+
"SR": "https://i.slow.pics/11ju3S94.webp"
53+
},
54+
{
55+
"type": "paired",
56+
"LR": "https://i.slow.pics/4RUqHkln.webp",
57+
"SR": "https://i.slow.pics/H1hh7YYK.webp"
58+
}
59+
]
60+
}

‎data/models/2x-AnimeSharpV2-RPLKSR-Soft.json

Lines changed: 60 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,60 @@
1+
{
2+
"name": "2x-AnimeSharpV2_RPLKSR_Soft",
3+
"author": "kim2091",
4+
"license": "CC-BY-NC-SA-4.0",
5+
"tags": [
6+
"anime",
7+
"compression-removal",
8+
"restoration"
9+
],
10+
"description": "GitHub Link: https://github.com/Kim2091/Kim2091-Models/releases/tag/2x-AnimeSharpV2_Set\n\nThis is my first anime model in years. Hopefully you guys can find a good use-case for it. Included are 4 models:\n\n1. RealPLKSR (Higher quality, slower)\n2. MoSR (Lower quality, faster)\n\nThere are Sharp and Soft versions of both\nWhen to use each:\n- __Sharp:__ For heavily degraded sources. Sharp models have issues depth of field but are best at removing artifacts \n- __Soft:__ For cleaner sources. Soft models preserve depth of field but may not remove other artifacts as well\n\n__Notes:__\n- MoSR doesn't work in chaiNNer currently\n- To use MoSR:\n 1. Use the ONNX version in tools like [VideoJaNai](<https://github.com/the-database/VideoJaNai>)\n 2. Update spandrel in the latest version of ComfyUI\n\nThe ONNX version may produce slightly different results than the .pth version. If you have issues, try the .pth model.",
11+
"date": "2024-10-05",
12+
"architecture": "realplksr",
13+
"size": null,
14+
"scale": 2,
15+
"inputChannels": 3,
16+
"outputChannels": 3,
17+
"resources": [
18+
{
19+
"platform": "pytorch",
20+
"type": "pth",
21+
"size": 29581666,
22+
"sha256": "a4c2ca131646db2603a061082df726212fa0bcb89031b7d7182a0b7f66e418d9",
23+
"urls": [
24+
"https://github.com/Kim2091/Kim2091-Models/releases/download/2x-AnimeSharpV2_Set/2x-AnimeSharpV2_RPLKSR_Soft.pth"
25+
]
26+
},
27+
{
28+
"platform": "onnx",
29+
"type": "onnx",
30+
"size": 15324988,
31+
"sha256": "e4177ec9a15dd7219bb4a8b9394ff16a8b97942265ee949ca6c46f90ce4d52e9",
32+
"urls": [
33+
"https://github.com/Kim2091/Kim2091-Models/releases/download/2x-AnimeSharpV2_Set/2x-AnimeSharpV2_RPLKSR_Soft_fp16.onnx"
34+
]
35+
}
36+
],
37+
"trainingIterations": 150000,
38+
"trainingBatchSize": 10,
39+
"trainingHRSize": 256,
40+
"trainingOTF": false,
41+
"dataset": "HFA2k Modified",
42+
"datasetSize": 3000,
43+
"images": [
44+
{
45+
"type": "paired",
46+
"LR": "https://i.slow.pics/7JZs9Otd.webp",
47+
"SR": "https://i.slow.pics/YPeC8Ilj.webp"
48+
},
49+
{
50+
"type": "paired",
51+
"LR": "https://i.slow.pics/mJZnrE4U.webp",
52+
"SR": "https://i.slow.pics/h7e3IIGY.webp"
53+
},
54+
{
55+
"type": "paired",
56+
"LR": "https://i.slow.pics/4RUqHkln.webp",
57+
"SR": "https://i.slow.pics/vRwR4fZK.webp"
58+
}
59+
]
60+
}

‎data/models/4x-PBRify-RPLKSRd-V3.json

Lines changed: 58 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,58 @@
1+
{
2+
"name": "4x-PBRify_RPLKSRd_V3",
3+
"author": "kim2091",
4+
"license": "CC0-1.0",
5+
"tags": [
6+
"compression-removal",
7+
"dds",
8+
"debanding",
9+
"dedither",
10+
"dehalo",
11+
"game-textures",
12+
"restoration"
13+
],
14+
"description": "PBRify Github: https://github.com/Kim2091/PBRify_Remix\n\nRelease Link: https://github.com/Kim2091/Kim2091-Models/releases/tag/4x-PBRify_RPLKSRd_V3\n\nThis update brings a new upscaling model, 4x-PBRify_RPLKSRd_V3. This model is roughly 8x faster than the current DAT2 model, while being *higher quality*. It produces far more natural detail, resolves lines and edges more smoothly, and cleans up compression artifacts better.\n\nAs a result of those improvements, PBR is also much improved. It tends to be clearer with less defined artifacts. \n\nHowever, this model is currently **only compatible with ComfyUI**. chaiNNer has not yet been updated to support this architecture.\n\n[More Comparisons](https://imgsli.com/Mjk5NjQ5)",
15+
"date": "2024-09-23",
16+
"architecture": "realplksr-dysample",
17+
"size": null,
18+
"scale": 4,
19+
"inputChannels": 3,
20+
"outputChannels": 3,
21+
"resources": [
22+
{
23+
"platform": "pytorch",
24+
"type": "pth",
25+
"size": 29717038,
26+
"sha256": "1325a08842ebfc18aea90e71594aa9b9e82b9c7e16321dccf5b757884a22daf1",
27+
"urls": [
28+
"https://github.com/Kim2091/Kim2091-Models/releases/download/4x-PBRify_RPLKSRd_V3/4x-PBRify_RPLKSRd_V3.pth"
29+
]
30+
}
31+
],
32+
"trainingIterations": 160000,
33+
"trainingBatchSize": 12,
34+
"trainingHRSize": 256,
35+
"trainingOTF": false,
36+
"dataset": "PolyHaven, FreePBR, ambientCG, UltraSharpV2",
37+
"datasetSize": 26000,
38+
"images": [
39+
{
40+
"type": "paired",
41+
"caption": "DAT2 vs RPLKSR",
42+
"LR": "https://imgsli.com/i/e4d8eaad-956c-4145-b827-4cfceecb9801.jpg",
43+
"SR": "https://imgsli.com/i/4d840f43-504b-4f15-bdf4-d548fdf7ec05.jpg"
44+
},
45+
{
46+
"type": "paired",
47+
"caption": "DAT2 vs RPLKSR 2",
48+
"LR": "https://imgsli.com/i/edaa03e0-4c1f-4a5d-ae30-e07cec74cc89.jpg",
49+
"SR": "https://imgsli.com/i/47d0cd27-2897-4749-b8c4-8aea10e49609.jpg"
50+
},
51+
{
52+
"type": "paired",
53+
"caption": "DAT2 vs RPLKSR 3",
54+
"LR": "https://imgsli.com/i/dae8bf21-6d48-4e97-a4a8-acd58e343918.jpg",
55+
"SR": "https://imgsli.com/i/747ffb24-619b-49a8-9720-6497a974549e.jpg"
56+
}
57+
]
58+
}

0 commit comments

Comments
 (0)