Almost all super resolution and interpolation models in a colab-notebook

[PatchWorKs]

We just found this very interesting repo (and relative colab):
Repository to use super resolution models and video frame interpolation models and also trying to speed them up with TensorRT. This repository contains the fastest inference code that you can find, at least I am trying to archive that. Not all codes can use TensorRT due to various reasons, but I try to add that if it works. Further model architectures are planned to be added later on.

Currently working networks:

ESRGAN with rlaphoenix/VSGAN (https://github.com/rlaphoenix/VSGAN) and HolyWu/vs-realesrgan (https://github.com/HolyWu/vs-realesrgan)
RealESRGAN / RealESERGANVideo with xinntao/Real-ESRGAN (https://github.com/xinntao/Real-ESRGAN) and rlaphoenix/VSGAN (https://github.com/rlaphoenix/VSGAN)
RealESRGAN ncnn with styler00dollar/realsr-ncnn-vulkan-python (https://github.com/styler00dollar/realsr-ncnn-vulkan-python) and media2x/realsr-ncnn-vulkan-python (https://github.com/media2x/realsr-ncnn-vulkan-python)
Rife4 with HolyWu/vs-rife (https://github.com/HolyWu/vs-rife/)
RIFE ncnn with styler00dollar/VapourSynth-RIFE-ncnn-Vulkan (https://github.com/styler00dollar/VapourSynth-RIFE-ncnn-Vulkan) and HomeOfVapourSynthEvolution/VapourSynth-RIFE-ncnn-Vulkan (https://github.com/HomeOfVapourSynthEvolution/VapourSynth-RIFE-ncnn-Vulkan)
SwinIR with HolyWu/vs-swinir (https://github.com/HolyWu/vs-swinir)
Sepconv (enhanced) with sniklaus/revisiting-sepconv (https://github.com/sniklaus/revisiting-sepconv/)
EGVSR with Thmen/EGVSR (https://github.com/Thmen/EGVSR) and HolyWu/vs-basicvsrpp (https://github.com/HolyWu/vs-basicvsrpp)
BasicVSR++ with HolyWu/vs-basicvsrpp (https://github.com/HolyWu/vs-basicvsrpp)
RealBasicVSR with ckkelvinchan/RealBasicVSR (https://github.com/HolyWu/vs-basicvsrpp)
RealCUGAN with bilibili/ailab (https://github.com/bilibili/ailab/blob/main/Real-CUGAN/README_EN.md)
FILM with google-research/frame-interpolation (https://github.com/google-research/frame-interpolation)
PAN with zhaohengyuan1/PAN (https://github.com/zhaohengyuan1/PAN)
IFRNet with ltkong218/IFRNet (https://github.com/ltkong218/IFRNet)
M2M with feinanshan/M2M_VFI (https://github.com/feinanshan/M2M_VFI)
IFUNet with 98mxr/IFUNet (https://github.com/98mxr/IFUNet/)
eisai with ShuhongChen/eisai-anime-interpolator (https://github.com/ShuhongChen/eisai-anime-interpolator)
SCUNet with cszn/SCUNet (https://github.com/cszn/SCUNet)
GMFupSS with 98mxr/GMFupSS (https://github.com/98mxr/GMFupSS)
ST-MFNet with danielism97/ST-MFNet (https://github.com/danielism97/ST-MFNet)
VapSR with zhoumumu/VapSR (https://github.com/zhoumumu/VapSR)


Also used:

TensorRT C++ inference with AmusementClub/vs-mlrt (https://github.com/AmusementClub/vs-mlrt)
ddfi with Mr-Z-2697/ddfi-rife (https://github.com/Mr-Z-2697/ddfi-rife) (auto dedup-duplication, not an arch)
nix with lucasew/nix-on-colab (https://github.com/lucasew/nix-on-colab)
custom ffmpeg with markus-perl/ffmpeg-build-script (https://github.com/markus-perl/ffmpeg-build-script)



VSGAN-tensorrt-docker (https://github.com/styler00dollar/VSGAN-tensorrt-docker)

PLEASE DON'T ABUSE (and, if you're able to, establish some kind of collaboration to optimize / evolve the project).

Enjoy and share results !

[SaurusX]

I've been using this for a few months now. It is the fastest way to use that suite of AI networks by far. TensorRT makes a hugemongous difference.

[Selur]

Would be nice to see a small comparison of all the methods. (Got no card with Tensor RT atm. :()