List of participants of MSU Video Super-Resolution Benchmark: Detail Restoration

Name Multi-frame Integrate temporal information by Training dataset Framework Year
D3Dnet Yes Deformable convolution Vimeo-90k PyTorch 2020
DBVSR Yes Optical flow REDS PyTorch 2020
DUF Yes Deformable convolution Unpublished TensorFlow 2018
DynaVSR-V Yes Vimeo-90k PyTorch 2020
DynaVSR-R Yes REDS PyTorch 2020
ESPCN No ImageNet PyTorch 2016
ESRGAN No PyTorch 2018
iSeeBetter Yes Recurrent architecture + optical flow PyTorch 2020
LGFN Yes Deformable convolution Vimeo-90k PyTorch 2020
RBPN Yes Recurrent architecture + optical flow Vimeo-90k PyTorch 2019
RRN Yes Recurrent architecture Vimeo-90k PyTorch 2020
Real-ESRGAN No PyTorch 2021
Real-ESRNet No PyTorch 2021
RealSR No DF2K, DPED PyTorch 2020
RSDN Yes Recurrent Vimeo-90k PyTorch 2020
SOF-VSR Yes Optical flow CDVL PyTorch 2020
TDAN Yes Deformable convolution Vimeo-90k PyTorch 2020
TGA Yes Temporal Group Attention Vimeo-90k PyTorch 2020
TMNet Yes Deformable convolution Vimeo-90k PyTorch 2021

D3Dnet

    Use a deformable 3D convolution to compensate the motion between frames implicitly.
    Added to the benchmark by MSU

DBVSR

    Estimate a motion blur for the particular input. Compensate the motion between frames explicitly.
    Added to the benchmark by MSU

DUF

    Two models: DUF-16L (16 layers), DUF-28L (28 layers)
    Use a deformable 3D convolution to compensate the motion between frames implicitly.
    Added to the benchmark by MSU

DynaVSR

    Two models: DynaVSR-R (trained on REDS), DynaVSR-V (trained on Vimeo-90k)
    Use meta-learning to estimate a degradation kernel for the particular input.
    DynaVSR can be applied to any VSR deep-learning model. For our benchmark, we used pretrained weights for model EDVR, which use Deformable convolution to align neighboring frames.
    Added to the benchmark by MSU

ESPCN

    Added to the benchmark by MSU

ESRGAN

    Added to the benchmark by MSU

iSeeBetter

    Added to the benchmark by MSU

LGFN

    Use deformable convolutions with decreased multi-dilation convolution units (DMDCUs) to align frames explicitly. Fuse features from local and global fusion modules.
    Added to the benchmark by MSU

RBPN

    Added to the benchmark by MSU

Real-ESRGAN

    Added to the benchmark by MSU

Real-ESRNet

    Added to the benchmark by MSU

RealSR

    Try to estimate degradation kernel and noise distribution for better visual quality.
    Added to the benchmark by MSU

RRN

    Two models: RRN-5L (five residual blocks), RRN-10L (ten residual blocks)
    Use recurrent strategy with sets of residual blocks to store information from previous frames.
    Added to the benchmark by MSU

RSDN

    It divides the input into structure and detail components which are fed to a recurrent unit composed of several proposed two-stream structure-detail blocks.
    Added to the benchmark by MSU

SOF-VSR

    Two models: SOF-VSR-BD (trained on gauss degradation type), SOF-VSR-BI (trained on bicubic degradation type)
    Compensate motion by high-resolution optical flow, estimated from the low-resolution one in a coarse-to-fine manner.
    Added to the benchmark by MSU

TDAN

    Use a deformable 3D convolution to compensate the motion between frames implicitly.
    Added to the benchmark by MSU

TGA

    The input sequence is reorganized into several groups of subsequences with different frame rates. The grouping allows to extract spatio-temporal information in a hierarchical manner, which is followed by an intra-group fusion module and inter-group fusion module.
    Added to the benchmark by MSU

TMNet

    Temporal Modulation Network was trained for Space-Time Video Super Resolution. The temporal information is integrated by deformable convolution with the multi-frame input.
    Added to the benchmark by the author, Gang Xu.
26 Apr 2021
See Also
MSU Video Upscalers Benchmark 2022
The most extensive comparison of video super-resolution (VSR) algorithms by subjective quality
MSU HDR Video Reconstruction Benchmark 2022
The most comprehensive comparison of HDR video reconstruction methods
Site structure