DEV Community

Željko Šević
Željko Šević

Posted on • Originally published at sevic.dev on

AI bulk image upscaler with Node.js

Image upscaling can be done using Real-ESRGAN, a super-resolution algorithm. Super-resolution is the process of increasing the resolution of the image.

Real-ESRGAN provides Linux, Windows and MacOS executable files and models for Intel/AMD/Nvidia GPUs.

The snippet below demonstrates bulk image upscaling with scale factor 4 and using the realesrgan-x4plus-anime model.

(async () => {
  const inputDirectory = path.resolve(path.join(__dirname, 'pictures'));
  const outputDirectory = path.resolve(
    path.join(__dirname, 'pictures_upscaled')
  );
  const modelsPath = path.resolve(path.join(__dirname, 'resources', 'models'));
  const execName = 'realesrgan-ncnn-vulkan';
  const execPath = path.resolve(
    path.join(__dirname, 'resources', getPlatform(), 'bin', execName)
  );
  const scaleFactor = 4;
  const modelName = 'realesrgan-x4plus-anime';

  if (!fs.existsSync(outputDirectory)) {
    await fs.promises.mkdir(outputDirectory, { recursive: true });
  }

  const commands = [
    '-i',
    inputDirectory,
    '-o',
    outputDirectory,
    '-s',
    scaleFactor,
    '-m',
    modelsPath,
    '-n',
    modelName
  ];

  const upscaler = spawn(execPath, commands, {
    cwd: undefined,
    detached: false
  });

  upscaler.stderr.on('data', (data) => {
    console.log(data.toString());
  });

  await timers.setTimeout(600 * 1000);
})();
Enter fullscreen mode Exit fullscreen mode

Demo

The demo with the mentioned example is available here.

Top comments (0)