Aura S R v2

fal

Introduction

AuraSR-V2 is a GAN-based super-resolution model designed for upscaling generated images. It is a variation of the GigaGAN architecture, specifically tailored for image-conditioned upscaling. The model is implemented in PyTorch and is available through the Hugging Face model hub.

Architecture

AuraSR-V2 leverages a GAN (Generative Adversarial Network) framework to enhance image resolution. It is based on the unofficial lucidrains/gigagan-pytorch repository and adapts the GigaGAN architecture for its super-resolution tasks. The model is optimized for image-to-image tasks and supports various tags including art and super-resolution.

Training

The model card does not provide specific training details for AuraSR-V2. However, it indicates that the implementation follows the strategies outlined in the GigaGAN paper, focusing on image-conditioned upscaling.

Guide: Running Locally

To run AuraSR-V2 locally, follow these steps:

  1. Install the Aura-SR Package:

    $ pip install aura-sr
    
  2. Load the Pre-trained Model:

    from aura_sr import AuraSR
    aura_sr = AuraSR.from_pretrained("fal/AuraSR-v2")
    
  3. Load and Upscale an Image:

    import requests
    from io import BytesIO
    from PIL import Image
    
    def load_image_from_url(url):
        response = requests.get(url)
        image_data = BytesIO(response.content)
        return Image.open(image_data)
    
    image = load_image_from_url("https://mingukkang.github.io/GigaGAN/static/images/iguana_output.jpg").resize((256, 256))
    upscaled_image = aura_sr.upscale_4x_overlapped(image)
    
  4. Hardware Recommendation:

    • For optimal performance, it is recommended to use a cloud GPU service such as AWS EC2 with GPU instances, Google Cloud Platform, or Azure.

License

AuraSR-V2 is licensed under the Apache 2.0 License. This allows for both personal and commercial use, distribution, and modification under the terms of the license.

More Related APIs in Image To Image