Real-ESRGAN

Real-ESRGAN is an extension to the powerful ESRGAN model and is able to generate high-resolution images with better perceptual quality than existing methods. The model has been tested across a variety of datasets and been shown to produce results that are comparable to or better than state-of-the-art methods.

Real-ESRGAN works by training on datasets of synthetic images which have been degraded in a similar fashion to real-world images. This allows the model to learn the patterns of degradation and how to recover them in order to generate high-resolution images with better perceptual quality.

Links

• GitHub
• Pre-trained models