Industry News
Enhancing Video Photorealism with AI
Using the Cityscapes training dataset, Stephan Richter (Intel Labs), Hassan Abu AlHaija (PCH Innovations / Intel Labs) and Vladlen Koltun (Intel Labs), have developed an approach to enhance the realism of synthetically generated images using AI.
The trio used a convolutional neural network to enhance the realism of real-time gameplay in GTA 5. The method is temporally stable and the modifications are both geometrically and semantically consistent with the original images.
Their method can also add reflections to windows, increase the fresnel effect on vehicle car paint, rebuild roads and even green parched grass hillsides.
Paper: https://arxiv.org/abs/2105.04619
Code and data: https://github.com/intel-isl/Photorea...
Project page: https://intel-isl.github.io/Photoreal...
The trio used a convolutional neural network to enhance the realism of real-time gameplay in GTA 5. The method is temporally stable and the modifications are both geometrically and semantically consistent with the original images.
Their method can also add reflections to windows, increase the fresnel effect on vehicle car paint, rebuild roads and even green parched grass hillsides.
Paper: https://arxiv.org/abs/2105.04619
Code and data: https://github.com/intel-isl/Photorea...
Project page: https://intel-isl.github.io/Photoreal...
Enhancing Photorealism Enhancement
Stephan R. Richter, Hassan Abu AlHaija, and Vladlen Koltun
Keynote for Eurographics 2021 - Vladlen Koltun: Further Towards Photorealism (May 2021)
See the GitHub video examples here: https://intel-isl.github.io/PhotorealismEnhancement/
You must be logged in to post a comment. Login here.
About this article
Using the Cityscapes training dataset, Stephan Richter (Intel Labs), Hassan Abu AlHaija (PCH Innovations / Intel Labs) and Vladlen Koltun (Intel Labs), have developed an approach to enhance the realism of synthetically generated images.
visibility1.93 k
favorite_border1
mode_comment0
