SwinIR
Swin Transformer image restoration for super-resolution, denoising, and JPEG cleanup.
About
SwinIR is the official PyTorch implementation of SwinIR: Image Restoration Using Swin Transformer, from Jingyun Liang and colleagues at ETH Zurich (ICCVW 2021). It was one of the first strong transformer baselines for low-level vision, using residual Swin Transformer blocks across shallow feature extraction, deep feature extraction, and reconstruction.
It covers classical, lightweight, and real-world image super-resolution, grayscale and color denoising, and JPEG compression artifact reduction, outperforming prior CNN methods by up to 0.14 to 0.45 dB while cutting parameters by up to 67 percent. Code and weights are Apache 2.0 licensed.
What it can do
- Image super-resolution (classical and real-world)
- Grayscale and color image denoising
- JPEG compression artifact reduction
- Apache 2.0 pretrained models
Setup
Clone the GitHub repo, install requirements, and download pretrained models from releases.
Prerequisites
Python, PyTorch, and a GPU for fast inference.
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 103; Trending score 68/100 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.