JingyunLiang/SwinIR

Health v2 · maintenance only

Adoption not part of the score

5580 stars · 665 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

Official PyTorch implementation of SwinIR, a Swin Transformer-based model for image restoration tasks including super-resolution, denoising, and JPEG compression artifact reduction. It provides pretrained models, training/evaluation code, and demos (Colab, Gradio, Replicate, mobile) achieving state-of-the-art results on low-level vision benchmarks.

Use cases

• upscale low-resolution images to high resolution
• remove noise from grayscale or color photos
• reduce JPEG compression artifacts in images
• restore real-world degraded photos with super-resolution
• run lightweight super-resolution on mobile devices
• deblock compressed images before further processing

When to choose

• you need state-of-the-art image super-resolution or denoising with pretrained models
• you want a transformer-based restoration baseline for research or comparison
• you need to clean up JPEG-compressed or noisy images in a Python pipeline
• you want a lightweight SR model that runs on mobile or via web demos

When to avoid

• you need video restoration rather than single-image processing
• you require real-time inference on CPU-constrained hardware
• you want a general-purpose image editor rather than a restoration model
• your task is high-level vision like detection or segmentation

Facets

library · maturity maintenance

image-processing computer-vision deep-learning machine-learning computer-vision image-processing deep-learning machine-learning python cross-platform super-resolution image-denoising jpeg-artifact-reduction image-deblocking swin-transformer vision-transformer pytorch low-level-vision pretrained-models image-restoration gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="JingyunLiang/SwinIR")