Skip to content

nfnt/sscale

Repository files navigation

sscale

Super-resolution image scaling using ESRGAN (Enhanced Super-Resolution Generative Adversarial Networks).

Overview

This tool upscales PNG and JPEG images by 4× using the ESRGAN model. It uses the 4x-ESRGAN.safetensors weights file converted from the PyTorch model at https://openmodeldb.info/models/4x-ESRGAN.

Example Usage

sscale input.jpg -o output.png

Performance

For GPU acceleration, build with the appropriate feature flag:

  • Linux: cargo build --release --features cuda or cargo build --release --features cudnn
  • macOS: cargo build --release --features metal

CPU inference is supported but significantly slower.

About

Super-resolution image scaling using ESRGAN

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published