Open Source Image and Video Restoration Toolbox for Super-resolution, Denoise, Deblurring, etc. Currently, it includes EDSR, RCAN, SRResNet, SRGAN, ESRGAN, EDVR, BasicVSR, SwinIR, ECBSR, etc. Also support StyleGAN2, DFDNet.

SwinIR: Image Restoration Using Swin Transformer (official repository)

Image Restoration Toolbox (PyTorch). Training and testing codes for DPIR, USRNet, DnCNN, FFDNet, SRMD, DPSR, BSRGAN, SwinIR

🌊 A Human-in-the-Loop workflow for creating HD images from text

Video Frame Interpolation & Super Resolution using NVIDIA's TensorRT & Tencent's NCNN inference, beautifully crafted and packaged into a single app

[ECCV] Swin2SR: SwinV2 Transformer for Compressed Image Super-Resolution and Restoration. Advances in Image Manipulation (AIM) workshop ECCV 2022. Try it out! over 3.3M runs https://replicate.com/mv-lab/swin2sr

A Flexible and Unified Image Restoration Framework (PyTorch), including state-of-the-art image restoration model. Such as NAFNet, Restormer, MPRNet, MIMO-UNet, SCUNet, SwinIR, HINet, etc. ⭐⭐⭐⭐⭐⭐

My VitePress website I made to visually compare the output I created with over 600 different AI upscaling models. Running under https://phhofm.github.io/upscale/

A High Quality (HD / 2K / 4K) Image Generation Using Stable Diffusion and Real-ESR / SwinIR /GFPGAN

SOTA image super-resolution in JavaScript/Wasm using an ONNX-ported SwinIR model

A repository of models trained by me!

No description available

experiment for reproducing SwinIR result

Fine-tune diffusion models on custom datasets and sample with text-conditioning using CLIP guidance combined with SwinIR for super resolution.

SwinIR function for VapourSynth

My self-trained AI models for photorealistic image upscaling and restoration. Built on ESRGAN, SwinIR, RealPLKSR and other architectures and trained on a manually curated high-end dataset (11,500+ real-world HD images). Models excel at restoring authentic textures and objects (skin, fabric, grit) and handle various types of degradation.

PyTorch implements `SwinIR: Image Restoration Using Swin Transformer` paper.

WindSR Dataset contains more than 22,000 pairs of HR/LR wind speed images, which are processed using the NASA's GEOS-5 Nature Run dataset. This dataset is useful for studying super-resolution for data collected using satellites rather natural RGB images.

Wrapper for "SwinIR: Image Restoration Using Swin Transformer", for easy usage as a package.

image super resolution using swin transformer v2

OCR(Korean)

Coronary dominance classification with ConvNext, SwinIR and ResNet 3D

A multistage text to image framework. Built from a inference-reduced set of min-dalle, glid-3-xl, and SwinIR.

No description available

JAX implementation of SwinIR model.

PyTorch implementation of our paper Road Sign Classification with Denoising Pipeline Approach

No description available

No description available

OCR(Korean)

No description available