ESRGAN
Open-source GAN model for 4x image super-resolution.
About
ESRGAN, short for Enhanced Super-Resolution Generative Adversarial Network, is an open-source model that upscales images 4x while generating plausible realistic detail. Its generator uses residual-in-residual dense blocks, with a discriminator pushing textures toward photorealism.
Published in 2018, it became the foundation for Real-ESRGAN and countless community upscaling tools, Colab notebooks, and local apps. Pretrained weights and implementations are freely available on GitHub.
What it can do
- 4x single-image super-resolution
- GAN-based perceptual detail generation
- Pretrained weights for photos and anime
- Foundation for Real-ESRGAN and community forks
Setup
Clone a GitHub implementation and run inference with the pretrained weights.
Prerequisites
Python with PyTorch; a GPU speeds up inference significantly.
Use cases
Works with
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 118; Trending score 72/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.