Add my new models (#438)

* Add ATD Arch * Add ATD 4x official pretrain * Create 4x-RealWebPhoto-v3-atd.json * Create 4x-RealWebPhoto-v4-dat2.json * Create 4x-Nomos8k-atd-jpg.json * Create 4x-RGT-S.json * Create 4x-RealWebPhoto-v2-rgt-s.json * Create 1x-Ludvae200.json * Fix 1x-Ludvae200 image links * 1x-Ludvae200 add pth file resource * Add 4x-TextureDAT2-otf * Add 4xTextures_GTAV_rgt-s * Add drct pretrains * Add span pretrains * Add 4xTextures_GTAV_rgt-s * Add 4xRealWebPhoto_v4_drct-l * Create 4x-Textures-GTAV-rgt-s-dither.json * 4x_realplksr_mssim_pretrain * Create 4x-realplksr-gan-pretrain.json * Create 4x-NomosWebPhoto-RealPLKSR.json * Create 4x-NomosWebPhoto-atd.json * Create 4x-NomosWebPhoto-esrgan.json * Create 4x-Nomos2-otf-esrgan.json * Create 4x-mssim-realplksr-dysample-pretrain.json * Create 4x-Nomos2-realplksr-dysample.json * Update 4x-Nomos2-otf-esrgan.json * Update 4x-NomosWebPhoto-atd.json
OpenModelDB · Jul 5, 2024 · 62a109d · 62a109d
1 parent 44f7c41
commit 62a109d
Show file tree

Hide file tree

Showing 12 changed files with 578 additions and 1 deletion.
diff --git a/data/architectures.json b/data/architectures.json
@@ -138,6 +138,14 @@
             "onnx"
         ]
     },
+    "realplksr": {
+        "name": "RealPLKSR",
+        "input": "image",
+        "compatiblePlatforms": [
+            "pytorch",
+            "onnx"
+        ]
+    },
     "rgt": {
         "name": "RGT",
         "input": "image",

diff --git a/data/models/4x-Nomos2-otf-esrgan.json b/data/models/4x-Nomos2-otf-esrgan.json
@@ -0,0 +1,99 @@
+{
+    "name": "4xNomos2_otf_esrgan",
+    "author": "helaman",
+    "license": "CC-BY-4.0",
+    "tags": [
+        "compression-removal",
+        "jpeg",
+        "photo",
+        "restoration"
+    ],
+    "description": "4xNomos2_otf_esrgan\n\nScale: 4  \nArchitecture: ESRGAN  \nArchitecture Option: esrgan  \n[Github Release Link](https://github.com/Phhofm/models/releases/tag/4xNomos2_otf_esrgan)  \n\nAuthor: Philip Hofmann  \nLicense: CC-BY-0.4  \nPurpose: Restoration  \nSubject: Photography  \nInput Type: Images  \nRelease Date: 22.06.2024  \n\nDataset: Nomos-v2  \nDataset Size: 6000  \nOTF (on the fly augmentations): Yes  \nPretrained Model: RealESRGAN_x4plus  \nIterations: 246'000  \nBatch Size: 8  \nGT Size: 256  \n\nDescription:  \n\n4x ESRGAN model for photography, trained using the Real-ESRGAN otf degradation pipeline.  \n\nShowcase:\n[Slow Pics 8 Examples](https://slow.pics/s/Y5wdyTPh?canvas-mode=fit-width&image-fit=contain)",
+    "date": "2024-06-22",
+    "architecture": "esrgan",
+    "size": [
+        "64nf",
+        "23nb"
+    ],
+    "scale": 4,
+    "inputChannels": 3,
+    "outputChannels": 3,
+    "resources": [
+        {
+            "platform": "pytorch",
+            "type": "pth",
+            "size": 33686030,
+            "sha256": "6c16d8b22b8468e0c0b766f1569e10a77d38fc7a1a929b804806dce84ddb85d3",
+            "urls": [
+                "https://github.com/Phhofm/models/releases/download/4xNomos2_otf_esrgan/4xNomos2_otf_esrgan.pth"
+            ]
+        },
+        {
+            "platform": "pytorch",
+            "type": "safetensors",
+            "size": 33467822,
+            "sha256": "12db878907ed3a52ee97de552dfc0ce7ffd38559ca9760a7803cb2d06a737055",
+            "urls": [
+                "https://github.com/Phhofm/models/releases/download/4xNomos2_otf_esrgan/4xNomos2_otf_esrgan.safetensors"
+            ]
+        },
+        {
+            "platform": "onnx",
+            "type": "onnx",
+            "size": 67003044,
+            "sha256": "21905712a7963b95bc7e6e0b506e2fc6f353742fa7bc3ccdb81c0ad47744cc8b",
+            "urls": [
+                "https://github.com/Phhofm/models/releases/download/4xNomos2_otf_esrgan/4xNomos2_otf_esrgan_fp32_opset17.onnx"
+            ]
+        }
+    ],
+    "trainingIterations": 246000,
+    "trainingBatchSize": 8,
+    "trainingHRSize": 256,
+    "trainingOTF": true,
+    "dataset": "Nomos-v2",
+    "datasetSize": 6000,
+    "pretrainedModelG": "4x-realesrgan-x4plus",
+    "images": [
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/s0aTG2Vz.png",
+            "SR": "https://i.slow.pics/TOL43AJ8.png"
+        },
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/556Unq6x.png",
+            "SR": "https://i.slow.pics/hF8hT44R.png"
+        },
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/lP3gidyo.png",
+            "SR": "https://i.slow.pics/nvQUv1Ah.png"
+        },
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/fBt2h5Lh.png",
+            "SR": "https://i.slow.pics/5TEHOFxf.png"
+        },
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/Xw0C7wiT.png",
+            "SR": "https://i.slow.pics/X4V4vmWk.png"
+        },
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/4XjROU2k.png",
+            "SR": "https://i.slow.pics/EMnCftls.png"
+        },
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/LhqBaoQa.png",
+            "SR": "https://i.slow.pics/6qr5sQZt.png"
+        },
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/5YsUu1mQ.png",
+            "SR": "https://i.slow.pics/gkbocFPC.png"
+        }
+    ]
+}
diff --git a/data/models/4x-Nomos2-realplksr-dysample.json b/data/models/4x-Nomos2-realplksr-dysample.json
@@ -0,0 +1,72 @@
+{
+    "name": "4xNomos2_realplksr_dysample",
+    "author": "helaman",
+    "license": "CC-BY-4.0",
+    "tags": [
+        "compression-removal",
+        "jpeg",
+        "photo",
+        "restoration"
+    ],
+    "description": "4xNomos2_realplksr_dysample  \nScale: 4  \nArchitecture: [RealPLKSR with Dysample](https://github.com/muslll/neosr/?tab=readme-ov-file#supported-archs)  \nArchitecture Option: [realplksr](https://github.com/muslll/neosr/blob/master/neosr/archs/realplksr_arch.py)  \n[Github Release](https://github.com/Phhofm/models/releases/tag/4xNomos2_realplksr_dysample)  \n\nAuthor: Philip Hofmann  \nLicense: CC-BY-0.4  \nPurpose: Pretrained  \nSubject: Photography  \nInput Type: Images  \nRelease Date: 30.06.2024  \n\nDataset: [nomosv2](https://github.com/muslll/neosr/?tab=readme-ov-file#-datasets)  \nDataset Size: 6000  \nOTF (on the fly augmentations): No  \nPretrained Model: [4xmssim_realplksr_dysample_pretrain](https://github.com/Phhofm/models/releases/tag/4xmssim_realplksr_dysample_pretrain)  \nIterations: 185'000  \nBatch Size: 8  \nGT Size: 256, 512  \n\nDescription: \nA Dysample RealPLKSR 4x upscaling model that was trained with / handles jpg compression down to 70 on the Nomosv2 dataset, preserves DoF.   \nBased on the [4xmssim_realplksr_dysample_pretrain](https://github.com/Phhofm/models/releases/tag/4xmssim_realplksr_dysample_pretrain) I released 3 days ago.  \nThis model affects / saturate colors, which can be counteracted a bit by using wavelet color fix, as used in these examples.  \n\nShowcase:\n[Slowpics](https://slow.pics/s/p3sMnd5l)",
+    "date": "2024-06-30",
+    "architecture": "realplksr",
+    "size": null,
+    "scale": 4,
+    "inputChannels": 3,
+    "outputChannels": 3,
+    "resources": [
+        {
+            "platform": "pytorch",
+            "type": "pth",
+            "size": 29715988,
+            "sha256": "96c8476ed2c42107628d10562b32ae89e7be3c2f7a53d29504823e9bafe8121d",
+            "urls": [
+                "https://github.com/Phhofm/models/releases/download/4xNomos2_realplksr_dysample/4xNomos2_realplksr_dysample.pth"
+            ]
+        },
+        {
+            "platform": "onnx",
+            "type": "onnx",
+            "size": 30298952,
+            "sha256": "f4da353c0a380aa2c53485f2479bfd7959cab071734075d96f53df3233976716",
+            "urls": [
+                "https://github.com/Phhofm/models/releases/download/4xNomos2_realplksr_dysample/4xNomos2_realplksr_dysample_256_fp32_fullyoptimized.onnx"
+            ]
+        }
+    ],
+    "trainingIterations": 185000,
+    "trainingBatchSize": 8,
+    "trainingHRSize": 512,
+    "trainingOTF": false,
+    "dataset": "Nomos-v2",
+    "datasetSize": 6000,
+    "pretrainedModelG": "4x-mssim-realplksr-dysample-pretrain",
+    "images": [
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/9ZnUcVDC.png",
+            "SR": "https://i.slow.pics/XJ5Tt9ja.png"
+        },
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/wdQgPTgv.png",
+            "SR": "https://i.slow.pics/1jspoDeR.png"
+        },
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/7NVM3NoB.png",
+            "SR": "https://i.slow.pics/481WUhuM.png"
+        },
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/4PZ6cOrU.png",
+            "SR": "https://i.slow.pics/xlPjpxUS.png"
+        },
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/SHPUrMz1.png",
+            "SR": "https://i.slow.pics/YVMbx6KC.png"
+        }
+    ]
+}
diff --git a/data/models/4x-NomosWebPhoto-RealPLKSR.json b/data/models/4x-NomosWebPhoto-RealPLKSR.json
@@ -0,0 +1,61 @@
+{
+    "name": "4xNomosWebPhoto_RealPLKSR",
+    "author": "helaman",
+    "license": "CC-BY-4.0",
+    "tags": [
+        "photo",
+        "restoration"
+    ],
+    "description": "4xNomosWebPhoto_RealPLKSR  \n\nScale: 4  \nArchitecture: RealPLKSR  \nArchitecture Option: realplksr  \n[Link to Github Release](https://github.com/Phhofm/models/releases/tag/4xNomosWebPhoto_RealPLKSR)  \n\nAuthor: Philip Hofmann  \nLicense: CC-BY-0.4  \nPurpose: Restoration  \nSubject: Photography  \nInput Type: Images  \nRelease Date: 28.05.2024  \n\nDataset: Nomos-v2  \nDataset Size: 6000  \nOTF (on the fly augmentations): No  \nPretrained Model: 4x_realplksr_gan_pretrain  \nIterations: 404'000, 445'000  \nBatch Size: 12, 4  \nGT Size: 128, 256, 512  \n\nDescription:\n\nshort: 4x RealPLKSR model for photography, trained with realistic noise, lens blur, jpg and webp re-compression.\n\nfull: My newest version of my RealWebPhoto series, this time I used the newly released Nomos-v2 dataset by musl.   \nI then made 12 different low resolution degraded folders, using kim's datasetdestroyer for scaling and compression, my ludvae200 model for realistic noise, and umzi's wtp_dataset_destroyer with its floating point lens blur implementation for better control.   \nI then mixed them together in a single lr folder and trained for 460'000 iters, checked the results, and upon kims suggestion of using interpolation, I tested and am releasing this interpolation between the checkpoints 404'000 and 445'000.  \n\nThis model has been trained on neosr using mixup, cutmix, resizemix, cutblur, nadam, unet, multisteplr, mssim, perceptual, gan, dists, ldl, ff, color and lumaloss, and interpolated using the current chaiNNer nightly version.  \n\nThis model took multiple retrainings and reworks of the dataset, until I am now satisfied enough with the quality to release this version.   \n\nFor more details on the whole process see the pdf file in the attachement.  \n\nI am also attaching the 404'000, 445'000 and 460'000 checkpoints for completeness.    \n\nPS in general degradation strengths have been reduced/adjusted in comparison to my previous RealWebPhoto models  \n\nShowcase:\n[Slow Pics 10 Examples](https://slow.pics/s/euvEv4hL)",
+    "date": "2024-05-28",
+    "architecture": "realplksr",
+    "size": [
+        "64dim",
+        "28nb",
+        "17ks",
+        "Real"
+    ],
+    "scale": 4,
+    "inputChannels": 3,
+    "outputChannels": 3,
+    "resources": [
+        {
+            "platform": "pytorch",
+            "type": "pth",
+            "size": 29683482,
+            "sha256": "a9db66c9b674c6a5025b6ef3bee71a57c33b8605d8a2de0980470f89002efbbe",
+            "urls": [
+                "https://github.com/Phhofm/models/releases/download/4xNomosWebPhoto_RealPLKSR/4xNomosWebPhoto_RealPLKSR.pth"
+            ]
+        }
+    ],
+    "trainingIterations": 445000,
+    "trainingBatchSize": 4,
+    "trainingHRSize": 512,
+    "trainingOTF": false,
+    "dataset": "Nomos-v2",
+    "datasetSize": 6000,
+    "pretrainedModelG": "4x-realplksr-gan-pretrain",
+    "images": [
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/ttBmEsyM.png",
+            "SR": "https://i.slow.pics/yHKf35eW.png"
+        },
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/68yeIooo.png",
+            "SR": "https://i.slow.pics/Z6vmBN69.png"
+        },
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/LUc7ikyl.png",
+            "SR": "https://i.slow.pics/IhI3wtGC.png"
+        },
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/mljAdTUv.png",
+            "SR": "https://i.slow.pics/XtfSCf8c.png"
+        }
+    ]
+}
diff --git a/data/models/4x-NomosWebPhoto-atd.json b/data/models/4x-NomosWebPhoto-atd.json
@@ -0,0 +1,69 @@
+{
+    "name": "4xNomosWebPhoto_atd",
+    "author": "helaman",
+    "license": "CC-BY-4.0",
+    "tags": [
+        "photo",
+        "restoration"
+    ],
+    "description": "4xNomosWebPhoto_atd\n\nScale: 4  \nArchitecture: [ATD](https://github.com/LabShuHangGU/Adaptive-Token-Dictionary)  \nArchitecture Option: atd  \n[Github Release Link](https://github.com/Phhofm/models/releases/tag/4xNomosWebPhoto_atd)  \n\n\nAuthor: Philip Hofmann   \nLicense: CC-BY-0.4  \nPurpose: Restoration  \nSubject: Photography  \nInput Type: Images  \nRelease Date: 07.06.2024  \n\nDataset: Nomos-v2  \nDataset Size: 6000  \nOTF (on the fly augmentations): No  \nPretrained Model: 003_ATD_SRx4_finetune.pth  \nIterations: 460'000  \nBatch Size: 6, 2  \nGT Size: 128, 192  \n\nDescription:  \n\n4x ATD model for photography, trained with realistic noise, lens blur, jpg and webp re-compression.   \nATD version of [4xNomosWebPhoto_RealPLKSR](https://github.com/Phhofm/models/releases/tag/4xNomosWebPhoto_RealPLKSR), trained on the same dataset and in the same way. For more information look into the [4xNomosWebPhoto_RealPLKSR](https://github.com/Phhofm/models/releases/tag/4xNomosWebPhoto_RealPLKSR) release, and the pdf file in its attachments.\n\nShowcase:\n[Slow Pics 18 Examples](https://slow.pics/s/GkS39XYi)",
+    "date": "2024-06-07",
+    "architecture": "atd",
+    "size": [
+        "210dim",
+        "16w",
+        "256cat"
+    ],
+    "scale": 4,
+    "inputChannels": 3,
+    "outputChannels": 3,
+    "resources": [
+        {
+            "platform": "pytorch",
+            "type": "pth",
+            "size": 81959074,
+            "sha256": "f1af8e13824e44dfaefd9954e4dd9a5abe916f40971ed323f63727735aa0983c",
+            "urls": [
+                "https://github.com/Phhofm/models/releases/download/4xNomosWebPhoto_atd/4xNomosWebPhoto_atd.pth"
+            ]
+        },
+        {
+            "platform": "pytorch",
+            "type": "safetensors",
+            "size": 81689540,
+            "sha256": "ca0e7a3ecbb042dad1695d4dd3b73cf355af6db822984b61ac363e42fac7dff1",
+            "urls": [
+                "/home/phhofm/Downloads/4xNomosWebPhoto_atd.safetensors"
+            ]
+        }
+    ],
+    "trainingIterations": 460000,
+    "trainingBatchSize": 2,
+    "trainingHRSize": 192,
+    "trainingOTF": false,
+    "dataset": "Nomos-v2",
+    "datasetSize": 6000,
+    "pretrainedModelG": "4x-003-ATD-SRx4-finetune",
+    "images": [
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/cJW6aPS0.png",
+            "SR": "https://i.slow.pics/CLWwbc3K.png"
+        },
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/OuZiJlE6.png",
+            "SR": "https://i.slow.pics/11QRFEdo.png"
+        },
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/Nls9WLwm.png",
+            "SR": "https://i.slow.pics/vbYgeTL8.png"
+        },
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/U06F3Fm2.png",
+            "SR": "https://i.slow.pics/RXBNdhut.png"
+        }
+    ]
+}
diff --git a/data/models/4x-NomosWebPhoto-esrgan.json b/data/models/4x-NomosWebPhoto-esrgan.json
@@ -0,0 +1,72 @@
+{
+    "name": "4xNomosWebPhoto_esrgan",
+    "author": "helaman",
+    "license": "CC-BY-4.0",
+    "tags": [
+        "photo",
+        "restoration"
+    ],
+    "description": "I simply wanted to release an ESRGAN model just because I had not trained one for quite a while and just wanted to revisit this older arch for the current series.\n\n4xNomosWebPhoto_esrgan\n\nScale: 4  \nArchitecture: ESRGAN  \nArchitecture Option: esrgan  \n[Github Release Link](https://github.com/Phhofm/models/releases/tag/4xNomosWebPhoto_esrgan)  \n\nAuthor: Philip Hofmann  \nLicense: CC-BY-0.4  \nPurpose: Restoration  \nSubject: Photography  \nInput Type: Images  \nRelease Date: 16.06.2024  \n\nDataset: Nomos-v2  \nDataset Size: 6000  \nOTF (on the fly augmentations): No  \nPretrained Model: RealESRGAN_x4plus  \nIterations: 210'000  \nBatch Size: 12  \nGT Size: 256  \n\nDescription:\n\n4x ESRGAN model for photography, trained with realistic noise, lens blur, jpg and webp re-compression.  \nESRGAN version of 4xNomosWebPhoto_RealPLKSR, trained on the same dataset and in a similiar way.  \nFor more information look into the 4xNomosWebPhoto_RealPLKSR release, and the pdf file in its attachments.",
+    "date": "2024-06-16",
+    "architecture": "esrgan",
+    "size": [
+        "64nf",
+        "23nb"
+    ],
+    "scale": 4,
+    "inputChannels": 3,
+    "outputChannels": 3,
+    "resources": [
+        {
+            "platform": "pytorch",
+            "type": "pth",
+            "size": 33688212,
+            "sha256": "131bb044af9fff0c748e460df1dce5aff018e2cd44340bbc6d97e683cdc0d491",
+            "urls": [
+                "https://github.com/Phhofm/models/releases/download/4xNomosWebPhoto_esrgan/4xNomosWebPhoto_esrgan.pth"
+            ]
+        },
+        {
+            "platform": "pytorch",
+            "type": "safetensors",
+            "size": 33467822,
+            "sha256": "2f49c3065db8aaeb64adb08e8879001fe2f1460df5655499707efedad8a3d556",
+            "urls": [
+                "https://github.com/Phhofm/models/releases/download/4xNomosWebPhoto_esrgan/4xNomosWebPhoto_esrgan.safetensors"
+            ]
+        },
+        {
+            "platform": "onnx",
+            "type": "onnx",
+            "size": 67003044,
+            "sha256": "e29f473727ee0e32416d3a2d2642568e0159e2608d67236193772fced8696121",
+            "urls": [
+                "https://github.com/Phhofm/models/releases/download/4xNomosWebPhoto_esrgan/4xNomosWebPhoto_esrgan_fp32_opset17.onnx"
+            ]
+        }
+    ],
+    "trainingIterations": 210000,
+    "trainingBatchSize": 12,
+    "trainingHRSize": 256,
+    "trainingOTF": false,
+    "dataset": "Nomos-v2",
+    "datasetSize": 6000,
+    "pretrainedModelG": "4x-realesrgan-x4plus",
+    "images": [
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/Q20xKISa.png",
+            "SR": "https://i.slow.pics/JFKE6GKT.png"
+        },
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/fY0zlrme.png",
+            "SR": "https://i.slow.pics/5yq29FLI.png"
+        },
+        {
+            "type": "paired",
+            "LR": "https://i.slow.pics/ORvP5MW6.png",
+            "SR": "https://i.slow.pics/fwrZBVdt.png"
+        }
+    ]
+}