nihui / waifu2x-ncnn-vulkan

waifu2x converter ncnn version, runs fast on intel / amd / nvidia / apple-silicon GPU with vulkan

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

waifu2x ncnn Vulkan

CI download

ncnn implementation of waifu2x converter. Runs fast on Intel / AMD / Nvidia / Apple-Silicon with Vulkan API.

waifu2x-ncnn-vulkan uses ncnn project as the universal neural network inference framework.


Download Windows/Linux/MacOS Executable for Intel/AMD/Nvidia GPU

This package includes all the binaries and models required. It is portable, so no CUDA or Caffe runtime environment is needed :)


Example Command

waifu2x-ncnn-vulkan.exe -i input.jpg -o output.png -n 2 -s 2

Full Usages

Usage: waifu2x-ncnn-vulkan -i infile -o outfile [options]...

  -h                   show this help
  -v                   verbose output
  -i input-path        input image path (jpg/png/webp) or directory
  -o output-path       output image path (jpg/png/webp) or directory
  -n noise-level       denoise level (-1/0/1/2/3, default=0)
  -s scale             upscale ratio (1/2/4/8/16/32, default=2)
  -t tile-size         tile size (>=32/0=auto, default=0) can be 0,0,0 for multi-gpu
  -m model-path        waifu2x model path (default=models-cunet)
  -g gpu-id            gpu device to use (-1=cpu, default=auto) can be 0,1,2 for multi-gpu
  -j load:proc:save    thread count for load/proc/save (default=1:2:2) can be 1:2,2,2:2 for multi-gpu
  -x                   enable tta mode
  -f format            output image format (jpg/png/webp, default=ext/png)
  • input-path and output-path accept either file path or directory path
  • noise-level = noise level, large value means strong denoise effect, -1 = no effect
  • scale = scale level, 1 = no scaling, 2 = upscale 2x
  • tile-size = tile size, use smaller value to reduce GPU memory usage, default selects automatically
  • load:proc:save = thread count for the three stages (image decoding + waifu2x upscaling + image encoding), using larger values may increase GPU usage and consume more GPU memory. You can tune this configuration with "4:4:4" for many small-size images, and "2:2:2" for large-size images. The default setting usually works fine for most situations. If you find that your GPU is hungry, try increasing thread count to achieve faster processing.
  • format = the format of the image to be output, png is better supported, however webp generally yields smaller file sizes, both are losslessly encoded

If you encounter a crash or error, try upgrading your GPU driver:

Build from Source

  1. Download and setup the Vulkan SDK from
  • For Linux distributions, you can either get the essential build requirements from package manager
dnf install vulkan-headers vulkan-loader-devel
apt-get install libvulkan-dev
pacman -S vulkan-headers vulkan-icd-loader
  1. Clone this project with all submodules
git clone
cd waifu2x-ncnn-vulkan
git submodule update --init --recursive
  1. Build with CMake
  • You can pass -DUSE_STATIC_MOLTENVK=ON option to avoid linking the vulkan loader library on MacOS
mkdir build
cd build
cmake ../src
cmake --build . -j 4

Speed Comparison with waifu2x-caffe-cui


  • Windows 10 1809
  • AMD R7-1700
  • Nvidia GTX-1070
  • Nvidia driver 419.67
  • CUDA 10.1.105
  • cuDNN 10.1
Measure-Command { waifu2x-ncnn-vulkan.exe -i input.png -o output.png -n 2 -s 2 -t [block size] -m [model dir] }
Measure-Command { waifu2x-caffe-cui.exe -t 0 --gpu 0 -b 1 -c [block size] -p cudnn --model_dir [model dir] -s 2 -n 2 -m noise_scale -i input.png -o output.png }


Image Size Target Size Block Size Total Time(s) GPU Memory(MB)
waifu2x-ncnn-vulkan 200x200 400x400 400/200/100 0.86/0.86/0.82 638/638/197
waifu2x-caffe-cui 200x200 400x400 400/200/100 2.54/2.39/2.36 3017/936/843
waifu2x-ncnn-vulkan 400x400 800x800 400/200/100 1.17/1.04/1.02 2430/638/197
waifu2x-caffe-cui 400x400 800x800 400/200/100 2.91/2.43/2.7 3202/1389/1178
waifu2x-ncnn-vulkan 1000x1000 2000x2000 400/200/100 2.35/2.26/2.46 2430/638/197
waifu2x-caffe-cui 1000x1000 2000x2000 400/200/100 4.04/3.79/4.35 3258/1582/1175
waifu2x-ncnn-vulkan 2000x2000 4000x4000 400/200/100 6.46/6.59/7.49 2430/686/213
waifu2x-caffe-cui 2000x2000 4000x4000 400/200/100 7.01/7.54/10.11 3258/1499/1200
waifu2x-ncnn-vulkan 4000x4000 8000x8000 400/200/100 22.78/23.78/27.61 2448/654/213
waifu2x-caffe-cui 4000x4000 8000x8000 400/200/100 18.45/21.85/31.82 3325/1652/1236


Image Size Target Size Block Size Total Time(s) GPU Memory(MB)
waifu2x-ncnn-vulkan 200x200 400x400 400/200/100 0.74/0.75/0.72 482/482/142
waifu2x-caffe-cui 200x200 400x400 400/200/100 2.04/1.99/1.99 995/546/459
waifu2x-ncnn-vulkan 400x400 800x800 400/200/100 0.95/0.83/0.81 1762/482/142
waifu2x-caffe-cui 400x400 800x800 400/200/100 2.08/2.12/2.11 995/546/459
waifu2x-ncnn-vulkan 1000x1000 2000x2000 400/200/100 1.52/1.41/1.44 1778/482/142
waifu2x-caffe-cui 1000x1000 2000x2000 400/200/100 2.72/2.60/2.68 1015/570/459
waifu2x-ncnn-vulkan 2000x2000 4000x4000 400/200/100 3.45/3.42/3.63 1778/482/142
waifu2x-caffe-cui 2000x2000 4000x4000 400/200/100 3.90/4.01/4.35 1015/521/462
waifu2x-ncnn-vulkan 4000x4000 8000x8000 400/200/100 11.16/11.29/12.07 1796/498/158
waifu2x-caffe-cui 4000x4000 8000x8000 400/200/100 9.24/9.81/11.16 995/546/436

Sample Images

Original Image


Upscale 2x with ImageMagick

convert origin.jpg -resize 200% output.png


Upscale 2x with ImageMagick Lanczo4 Filter

convert origin.jpg -filter Lanczos -resize 200% output.png


Upscale 2x with waifu2x noise=2 scale=2

waifu2x-ncnn-vulkan.exe -i origin.jpg -o output.png -n 2 -s 2


Original waifu2x Project

Other Open-Source Code Used


waifu2x converter ncnn version, runs fast on intel / amd / nvidia / apple-silicon GPU with vulkan

License:MIT License


Language:C 82.7%Language:C++ 14.2%Language:CMake 3.1%