[MODEL ADD REQUEST] 4x-PBRify_UpscalerSIR-M_V2

[Kim2091]

{
  "name": "4x-PBRify_UpscalerSIR-M_V2",
  "author": "kim2091",
  "license": "CC0-1.0",
  "tags": [
    "dedither",
    "game-textures",
    "restoration"
  ],
  "description": "This is part of my [PBRify_Remix](<https://github.com/Kim2091/PBRify_Remix>) project. This is a much more capable model based on SwinIR Medium, which should strike a balance between capacity for learning + inference speed. It appears to have done so :)",
  "date": "2024-05-19",
  "architecture": "swinir",
  "size": [
    "Medium"
  ],
  "scale": 4,
  "inputChannels": 3,
  "outputChannels": 3,
  "resources": [
    {
      "urls": [
        "https://github.com/Kim2091/Kim2091-Models/releases/download/4x-PBRify_UpscalerSIR-M_V2/4x-PBRify_UpscalerSIR-M_V2.pth"
      ],
      "sha256": "65f5706973c4a66321b26681f3de3583c73b807dd97dd00fc69dfdb747396fd2",
      "size": 59606776,
      "platform": "pytorch",
      "type": "pth"
    }
  ],
  "images": [
    {
      "type": "paired",
      "LR": "https://i.slow.pics/xJmIuxId.png",
      "SR": "https://i.slow.pics/uEF2YWSy.png",
      "caption": "Input vs Output (RTX Remix)"
    },
    {
      "type": "paired",
      "LR": "https://i.slow.pics/pOpqMtIv.png",
      "SR": "https://i.slow.pics/KExMUh07.png"
    },
    {
      "type": "paired",
      "LR": "https://i.slow.pics/p1BX66df.png",
      "SR": "https://i.slow.pics/eDZkR2gl.png"
    },
    {
      "type": "paired",
      "LR": "https://i.slow.pics/YyrZBxVe.png",
      "SR": "https://i.slow.pics/O38UCmyY.png",
      "caption": "Old vs New SR Model"
    }
  ],
  "dataset": "ambientCG + PolyHaven + UltraSharpV2",
  "datasetSize": 26000,
  "trainingIterations": 400000,
  "trainingBatchSize": 8,
  "trainingHRSize": 256,
  "trainingOTF": false
}