TasksSotADatasetsPapersMethodsSubmitAbout
Papers With Code 2

A community resource for machine learning research: papers, code, benchmarks, and state-of-the-art results.

Explore

Notable BenchmarksAll SotADatasetsPapersMethods

Community

Submit ResultsAbout

Data sourced from the PWC Archive (CC-BY-SA 4.0). Built by the community, for the community.

Models/SwinFIR

SwinFIR

Reported on 112 benchmarks across 4 tasks · 1 paper

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Computer Vision56 results

  • Image Super-ResolutiononBSD100 - 2x upscaling
    PSNR· 2022-08-24
    32.64
    best: 33.12 (WaveMixSR-V2)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononBSD100 - 2x upscaling
    SSIM· 2022-08-24
    0.9054
    best: 0.9326 (WaveMixSR-V2)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononSet14 - 3x upscaling
    PSNR· 2022-08-24
    31.24
    best: 31.55 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononSet14 - 3x upscaling
    SSIM· 2022-08-24
    0.8566
    best: 0.8616 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononSet14 - 2x upscaling
    PSNR· 2022-08-24
    34.93
    best: 35.36 (DRCT-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononSet14 - 2x upscaling
    SSIM· 2022-08-24
    0.9276
    best: 0.9311 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononSet14 - 4x upscaling
    PSNR· 2022-08-24
    29.36
    best: 29.54 (DRCT-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononSet14 - 4x upscaling
    SSIM· 2022-08-24
    0.7993
    best: 0.894 (Edge-informed SR)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononSet5 - 3x upscaling
    PSNR· 2022-08-24
    35.15
    best: 35.35 (HMA†)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononSet5 - 3x upscaling
    SSIM· 2022-08-24
    0.933
    best: 0.938 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononManga109 - 4x upscaling
    PSNR· 2022-08-24
    32.83
    best: 33.19 (HMA†)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononManga109 - 4x upscaling
    SSIM· 2022-08-24
    0.9314
    best: 0.9366 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononUrban100 - 2x upscaling
    PSNR· 2022-08-24
    34.57
    best: 35.24 (HMA†)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononUrban100 - 2x upscaling
    SSIM· 2022-08-24
    0.9473
    best: 0.9516 (DRCT-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononManga109 - 3x upscaling
    PSNR· 2022-08-24
    35.77
    best: 36.12 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononManga109 - 3x upscaling
    SSIM· 2022-08-24
    0.9563
    best: 0.9588 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononSet5 - 2x upscaling
    PSNR· 2022-08-24
    38.65
    best: 39.14 (DRCT-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononSet5 - 2x upscaling
    SSIM· 2022-08-24
    0.9633
    best: 0.9663 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononManga109 - 2x upscaling
    PSNR· 2022-08-24
    40.61
    best: 41.22 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononManga109 - 2x upscaling
    SSIM· 2022-08-24
    0.9816
    best: 0.9846 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononUrban100 - 4x upscaling
    PSNR· 2022-08-24
    28.12
    best: 28.72 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononUrban100 - 4x upscaling
    SSIM· 2022-08-24
    0.8393
    best: 0.9481 (SPSR)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononUrban100 - 3x upscaling
    PSNR· 2022-08-24
    30.43
    best: 31.07 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononUrban100 - 3x upscaling
    SSIM· 2022-08-24
    0.8913
    best: 0.902 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononBSD100 - 4x upscaling
    PSNR· 2022-08-24
    28.03
    best: 28.16 (DRCT-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononBSD100 - 4x upscaling
    SSIM· 2022-08-24
    0.752
    best: 0.851 (Edge-informed SR)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononBSD100 - 3x upscaling
    PSNR· 2022-08-24
    29.55
    best: 29.67 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Image Super-ResolutiononBSD100 - 3x upscaling
    SSIM· 2022-08-24
    0.8169
    best: 0.8256 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononBSD100 - 2x upscaling
    PSNR· 2022-08-24
    32.64
    best: 33.12 (WaveMixSR-V2)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononBSD100 - 2x upscaling
    SSIM· 2022-08-24
    0.9054
    best: 0.9326 (WaveMixSR-V2)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononSet14 - 3x upscaling
    PSNR· 2022-08-24
    31.24
    best: 31.55 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononSet14 - 3x upscaling
    SSIM· 2022-08-24
    0.8566
    best: 0.8616 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononSet14 - 2x upscaling
    PSNR· 2022-08-24
    34.93
    best: 35.36 (DRCT-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononSet14 - 2x upscaling
    SSIM· 2022-08-24
    0.9276
    best: 0.9311 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononSet14 - 4x upscaling
    PSNR· 2022-08-24
    29.36
    best: 29.54 (DRCT-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononSet14 - 4x upscaling
    SSIM· 2022-08-24
    0.7993
    best: 0.894 (Edge-informed SR)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononSet5 - 3x upscaling
    PSNR· 2022-08-24
    35.15
    best: 35.35 (HMA†)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononSet5 - 3x upscaling
    SSIM· 2022-08-24
    0.933
    best: 0.938 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononManga109 - 4x upscaling
    PSNR· 2022-08-24
    32.83
    best: 33.19 (HMA†)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononManga109 - 4x upscaling
    SSIM· 2022-08-24
    0.9314
    best: 0.9366 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononUrban100 - 2x upscaling
    PSNR· 2022-08-24
    34.57
    best: 35.24 (HMA†)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononUrban100 - 2x upscaling
    SSIM· 2022-08-24
    0.9473
    best: 0.9516 (DRCT-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononManga109 - 3x upscaling
    PSNR· 2022-08-24
    35.77
    best: 36.12 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononManga109 - 3x upscaling
    SSIM· 2022-08-24
    0.9563
    best: 0.9588 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononSet5 - 2x upscaling
    PSNR· 2022-08-24
    38.65
    best: 39.14 (DRCT-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononSet5 - 2x upscaling
    SSIM· 2022-08-24
    0.9633
    best: 0.9663 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononManga109 - 2x upscaling
    PSNR· 2022-08-24
    40.61
    best: 41.22 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononManga109 - 2x upscaling
    SSIM· 2022-08-24
    0.9816
    best: 0.9846 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononUrban100 - 4x upscaling
    PSNR· 2022-08-24
    28.12
    best: 28.72 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononUrban100 - 4x upscaling
    SSIM· 2022-08-24
    0.8393
    best: 0.9481 (SPSR)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononUrban100 - 3x upscaling
    PSNR· 2022-08-24
    30.43
    best: 31.07 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononUrban100 - 3x upscaling
    SSIM· 2022-08-24
    0.8913
    best: 0.902 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononBSD100 - 4x upscaling
    PSNR· 2022-08-24
    28.03
    best: 28.16 (DRCT-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononBSD100 - 4x upscaling
    SSIM· 2022-08-24
    0.752
    best: 0.851 (Edge-informed SR)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononBSD100 - 3x upscaling
    PSNR· 2022-08-24
    29.55
    best: 29.67 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 3D Object Super-ResolutiononBSD100 - 3x upscaling
    SSIM· 2022-08-24
    0.8169
    best: 0.8256 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247

Graphs28 results

  • Super-ResolutiononBSD100 - 2x upscaling
    PSNR· 2022-08-24
    32.64
    best: 33.12 (WaveMixSR-V2)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononBSD100 - 2x upscaling
    SSIM· 2022-08-24
    0.9054
    best: 0.9326 (WaveMixSR-V2)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononSet14 - 3x upscaling
    PSNR· 2022-08-24
    31.24
    best: 31.55 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononSet14 - 3x upscaling
    SSIM· 2022-08-24
    0.8566
    best: 0.8616 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononSet14 - 2x upscaling
    PSNR· 2022-08-24
    34.93
    best: 35.36 (DRCT-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononSet14 - 2x upscaling
    SSIM· 2022-08-24
    0.9276
    best: 0.9311 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononSet14 - 4x upscaling
    PSNR· 2022-08-24
    29.36
    best: 29.54 (DRCT-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononSet14 - 4x upscaling
    SSIM· 2022-08-24
    0.7993
    best: 0.894 (Edge-informed SR)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononSet5 - 3x upscaling
    PSNR· 2022-08-24
    35.15
    best: 35.35 (HMA†)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononSet5 - 3x upscaling
    SSIM· 2022-08-24
    0.933
    best: 0.938 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononManga109 - 4x upscaling
    PSNR· 2022-08-24
    32.83
    best: 33.19 (HMA†)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononManga109 - 4x upscaling
    SSIM· 2022-08-24
    0.9314
    best: 0.9366 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononUrban100 - 2x upscaling
    PSNR· 2022-08-24
    34.57
    best: 35.24 (HMA†)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononUrban100 - 2x upscaling
    SSIM· 2022-08-24
    0.9473
    best: 0.9516 (DRCT-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononManga109 - 3x upscaling
    PSNR· 2022-08-24
    35.77
    best: 36.12 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononManga109 - 3x upscaling
    SSIM· 2022-08-24
    0.9563
    best: 0.9588 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononSet5 - 2x upscaling
    PSNR· 2022-08-24
    38.65
    best: 39.14 (DRCT-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononSet5 - 2x upscaling
    SSIM· 2022-08-24
    0.9633
    best: 0.9663 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononManga109 - 2x upscaling
    PSNR· 2022-08-24
    40.61
    best: 41.22 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononManga109 - 2x upscaling
    SSIM· 2022-08-24
    0.9816
    best: 0.9846 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononUrban100 - 4x upscaling
    PSNR· 2022-08-24
    28.12
    best: 28.72 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononUrban100 - 4x upscaling
    SSIM· 2022-08-24
    0.8393
    best: 0.9481 (SPSR)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononUrban100 - 3x upscaling
    PSNR· 2022-08-24
    30.43
    best: 31.07 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononUrban100 - 3x upscaling
    SSIM· 2022-08-24
    0.8913
    best: 0.902 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononBSD100 - 4x upscaling
    PSNR· 2022-08-24
    28.03
    best: 28.16 (DRCT-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononBSD100 - 4x upscaling
    SSIM· 2022-08-24
    0.752
    best: 0.851 (Edge-informed SR)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononBSD100 - 3x upscaling
    PSNR· 2022-08-24
    29.55
    best: 29.67 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • Super-ResolutiononBSD100 - 3x upscaling
    SSIM· 2022-08-24
    0.8169
    best: 0.8256 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247

Methodology28 results

  • 16konBSD100 - 2x upscaling
    PSNR· 2022-08-24
    32.64
    best: 33.12 (WaveMixSR-V2)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konBSD100 - 2x upscaling
    SSIM· 2022-08-24
    0.9054
    best: 0.9326 (WaveMixSR-V2)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konSet14 - 3x upscaling
    PSNR· 2022-08-24
    31.24
    best: 31.55 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konSet14 - 3x upscaling
    SSIM· 2022-08-24
    0.8566
    best: 0.8616 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konSet14 - 2x upscaling
    PSNR· 2022-08-24
    34.93
    best: 35.36 (DRCT-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konSet14 - 2x upscaling
    SSIM· 2022-08-24
    0.9276
    best: 0.9311 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konSet14 - 4x upscaling
    PSNR· 2022-08-24
    29.36
    best: 29.54 (DRCT-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konSet14 - 4x upscaling
    SSIM· 2022-08-24
    0.7993
    best: 0.894 (Edge-informed SR)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konSet5 - 3x upscaling
    PSNR· 2022-08-24
    35.15
    best: 35.35 (HMA†)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konSet5 - 3x upscaling
    SSIM· 2022-08-24
    0.933
    best: 0.938 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konManga109 - 4x upscaling
    PSNR· 2022-08-24
    32.83
    best: 33.19 (HMA†)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konManga109 - 4x upscaling
    SSIM· 2022-08-24
    0.9314
    best: 0.9366 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konUrban100 - 2x upscaling
    PSNR· 2022-08-24
    34.57
    best: 35.24 (HMA†)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konUrban100 - 2x upscaling
    SSIM· 2022-08-24
    0.9473
    best: 0.9516 (DRCT-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konManga109 - 3x upscaling
    PSNR· 2022-08-24
    35.77
    best: 36.12 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konManga109 - 3x upscaling
    SSIM· 2022-08-24
    0.9563
    best: 0.9588 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konSet5 - 2x upscaling
    PSNR· 2022-08-24
    38.65
    best: 39.14 (DRCT-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konSet5 - 2x upscaling
    SSIM· 2022-08-24
    0.9633
    best: 0.9663 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konManga109 - 2x upscaling
    PSNR· 2022-08-24
    40.61
    best: 41.22 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konManga109 - 2x upscaling
    SSIM· 2022-08-24
    0.9816
    best: 0.9846 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konUrban100 - 4x upscaling
    PSNR· 2022-08-24
    28.12
    best: 28.72 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konUrban100 - 4x upscaling
    SSIM· 2022-08-24
    0.8393
    best: 0.9481 (SPSR)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konUrban100 - 3x upscaling
    PSNR· 2022-08-24
    30.43
    best: 31.07 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konUrban100 - 3x upscaling
    SSIM· 2022-08-24
    0.8913
    best: 0.902 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konBSD100 - 4x upscaling
    PSNR· 2022-08-24
    28.03
    best: 28.16 (DRCT-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konBSD100 - 4x upscaling
    SSIM· 2022-08-24
    0.752
    best: 0.851 (Edge-informed SR)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konBSD100 - 3x upscaling
    PSNR· 2022-08-24
    29.55
    best: 29.67 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247
  • 16konBSD100 - 3x upscaling
    SSIM· 2022-08-24
    0.8169
    best: 0.8256 (Hi-IR-L)
    SwinFIR: Revisiting the SwinIR with Fast Fourier Convolution and Improved Training for Image Super-ResolutionarXiv:2208.11247