AI Research

Enhanced Super-Resolution GAN Remasters Max Payne

Researchers using enhanced super-resolution technology are giving classic video games of the past incredible, texture-rich visual makeovers. The team has released ‘remastered’ versions of Return to Castle Wolfenstein, Doom, The Elder Scrolls III: Morrowind, and most recently — a visually enhanced version of 2001 third-person shooter game Max Payne.

Researchers using enhanced super-resolution technology are giving classic video games of the past incredible, texture-rich visual makeovers. The team has released ‘remastered’ versions of Return to Castle Wolfenstein, Doom, The Elder Scrolls III: Morrowind, and most recently — a visually enhanced version of 2001 third-person shooter game Max Payne.

Image super-resolution refers to the reconstruction of corresponding high-resolution images from observed low-resolution images. These new reconstructions not only make images sharper and more detailed, but also reconstruct important missing elements using the advanced semantic information of the images.

The improved textures in the remastered games were all built using ESRGAN (enhanced super-resolution generative adversarial networks). The technique adds real-world detail to improve the realism of low-resolution images while also maintaining their original textures. Video game purists in particular appreciate that a classic game’s original textures don’t need to be replaced, which preserves their distinctive artistic styles.

Compared to traditional super-resolution (SR) methods, ESRGAN produces images with higher fidelity and realism by teaching the discriminator to judge “whether one image is more realistic than the other” rather than “whether one image is real or fake.” Researchers also employed residual scaling and smaller initialization, relativistic discriminators and perceptual loss functions.

The ESRGAN model outperforms previous SR methods in perceptual quality, and bested the PIRM-SR Challenge perceptual index. The team comprises researchers from from the Chinese University of Hong Kong, Shenzhen Institute of Advanced Technology, University of Chinese Academy of Sciences and Nanyang Technological University.

The team’s original paper ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks is on arXiv.


Author: Herin Zhao | Editor: Michael Sarazen

0 comments on “Enhanced Super-Resolution GAN Remasters Max Payne

Leave a Reply

Your email address will not be published. Required fields are marked *

%d bloggers like this: