RENVI - Real-ESRGAN ncnn Vulkan Improved

RENVI (or Real-ESRGAN ncnn Vulkan Improved) is sligthly improved version of Real-ESRGAN ncnn Vulkan with additional features like piping and fast PNG output.

Additional Features

• Improved performance: Uses libpng with low compression level for faster PNG writing.
• Piping: Supports reading from stdin and writing to stdout, allowing easy integration into pipelines.

Important

Only PNG format supports piping currently. JPEG and WebP formats are not supported for stdin/stdout operations. See the example usage below.

Installation

You can download the latest release from the Releases page or build it from source:

Dependencies

Arch Linux

sudo pacman -S base-devel cmake libpng glslang vulkan-headers vulkan-icd-loader openmp

Debian / Ubuntu

sudo apt install build-essential cmake libpng-dev glslang-dev libvulkan-dev libomp-dev

Fedora / RHEL

sudo dnf install gcc gcc-c++ make cmake libpng-devel glslang vulkan-headers vulkan-loader-devel libomp-devel

git clone --recurse-submodules -j8 git@github.com:ONdraid/Real-ESRGAN-ncnn-vulkan-improved.git

cd Real-ESRGAN-ncnn-vulkan-improved

mkdir build && cd build

cmake ../src

cmake --build . -j 4

Usages

Example Usage

You can create a full ffmpeg pipeline to upscale a video!

ffmpeg -hide_banner -loglevel error -i input.mkv -f image2pipe -vcodec png - \
| ./realesrgan-ncnn-vulkan-improved -n realesr-animevideov3 -s 2 \
| ffmpeg -f image2pipe -vcodec png -i - output.mkv

Tip

Ffmpeg defaults to 25 fps, use -r option to set the desired frame rate for the output video.

Full usage

Usage: realesrgan-ncnn-vulkan-improved [options]...

  -h                   show this help
  -i input-path        input image path (jpg/png/webp) or directory (reads from stdin if not provided)
  -o output-path       output image path (jpg/png/webp) or directory (outputs to stdout if not provided)
  -s scale             upscale ratio (can be 2, 3, 4. default=4)
  -t tile-size         tile size (>=32/0=auto, default=0) can be 0,0,0 for multi-gpu
  -m model-path        folder path to the pre-trained models. default=models
  -n model-name        model name (default=realesr-animevideov3, can be realesr-animevideov3 | realesrgan-x4plus | realesrgan-x4plus-anime | realesrnet-x4plus)
  -g gpu-id            gpu device to use (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)
  -v                   verbose output

Note

When utilizing stdin/stdout, load/save threads are fixed at 1:1, only proc threads can be adjusted in this configuration.

Star History