https://github.com/zsxkib/cog-aura-sr
AuraSR: GAN-based Super-Resolution for real-world
https://github.com/zsxkib/cog-aura-sr
Last synced: about 2 months ago
JSON representation
AuraSR: GAN-based Super-Resolution for real-world
- Host: GitHub
- URL: https://github.com/zsxkib/cog-aura-sr
- Owner: zsxkib
- License: mit
- Created: 2024-06-27T12:06:23.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2024-06-27T21:27:28.000Z (over 1 year ago)
- Last Synced: 2024-12-16T11:05:57.173Z (10 months ago)
- Language: Python
- Size: 12.7 KB
- Stars: 4
- Watchers: 3
- Forks: 0
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# AuraSR: GAN Super-Resolution for Images 🖼️
[](https://replicate.com/zsxkib/aura-sr)
AuraSR is a powerful tool that makes images bigger and clearer. It's based on the [GigaGAN](https://mingukkang.github.io/GigaGAN/) idea and works great for certain types of images.
## Quick Start 🚀
Use Cog to make your image bigger:
```bash
cog predict -i image=@your_image.png -i scale_factor=4
```
## What It Does 🎨
- Makes PNG, lossless WebP, and high-quality JPEG XL (90+) images bigger and clearer
- Can make images 2, 4, 8, 16, or 32 times bigger
- Works fast and can handle different sized jobs
## Things to Know ⚠️
AuraSR is great, but it has some limits:
1. It works best with PNG, lossless WebP, and high-quality JPEG XL (90+) images.
2. It doesn't like images that have been squeezed too much (compressed).
3. It can't fix mistakes in images.
4. It's best for making AI-generated images or very high-quality photos bigger.
## What You Need 📋
- Python 3.7 or newer
- Cog
## How to Use It 🛠️
### Simple Way
```bash
cog predict -i image=@your_image.png -i scale_factor=4
```
### Advanced Way
```bash
cog predict -i image=@your_image.png -i scale_factor=8 -i max_batch_size=4
```
## Options 🔧
- `image`: The picture you want to make bigger
- `scale_factor`: How much bigger you want to make it (2, 4, 8, 16, or 32)
- `max_batch_size`: Controls how fast it works (default is 1, increase if you have a powerful computer)
## Thank You 🙌
- [GAN-based Super-Resolution](https://github.com/fal-ai/aura-sr/tree/main) for real-world images, a variation of the GigaGAN paper for image-conditioned upscaling. PyTorch implementation is based on the unofficial [lucidrains/gigagan-pytorch](https://github.com/lucidrains/gigagan-pytorch) repository.
## Let's Talk 🐦
Have questions? Follow me on Twitter [@zsakib_](https://twitter.com/zsakib_) and let's chat!