Enroll Course

100% Online Study
Web & Video Lectures
Earn Diploma Certificate
Access to Job Openings
Access to CV Builder



Online Certification Courses

Intel AI Has The Potential To Make Games More Realistic In The Future

Development, Programming, Artificial Intelligence. 

Intel's AI has the potential to make games more realistic in the future

In May, there was a report on Intel's use of artificial intelligence to enhance the visuals in Grand Theft Auto V. Now, the company is discussing how it might be used in the future.

This future may include the incorporation of artificial intelligence to enhance the level of detail in future games. This could take the form of post-processing, baking it directly into the game, or even altering the way visual assets are created.

GTA V has been modded a few times since its release, but even the best mods cannot make it look as realistic as Intel's machine learning work.

The game's visuals have been significantly improved as a result of a machine learning project being conducted by a team at Intel Labs.

The project, dubbed "Enhancing Photorealism Enhancement," took scenes from Grand Theft Auto and enhanced the visuals to make them more realistic, with astounding results.

As the clips and images demonstrate, the work of Stephan R. Richter, Hassan Abu Alhaija, and Vladlen Kolten transformed the visuals with more realistic washed-out lighting, convincing real-world-styled reflections, and much more. 

 

How to achieve this look

To achieve this look, the researchers trained a neural network on Cityscapes Datasets and taught it real-world views as seen through car-mounted cameras. This training was then combined with logic from the game's engine, which sends information about the depths of objects the user is viewing and the lighting situation, to create a convincing view.

Currently, it is not ready for widespread use because the intelligence is not applied in real-time. Stephan Richter, the project's co-author, explained that "...it takes half a second to render a single frame...but the code is not optimized, so I could imagine it running much faster."

Racing and driving games may be ideal for utilizing this technique and the intelligence behind it, as it is relatively easy to obtain footage that corresponds to what is happening in-game and what is happening in the real world. However, the researchers assert that this is not the only possibility and that the possibilities are limitless.

Although the team believes it will be several years before this technique is used in real-time in games, the future is certainly exciting.

The detailed paper contains a great deal more information about how this works. Though we're unlikely to see this as an official mod for GTA V, it's impressive to see the results of this machine learning and the potential impact on future AI upscaling of games.

 

Courses and Certification

Artificial Intelligence Course and Certificate

Artificial Neural Network Course and Certificate

Machine Learning with Python Course and Certificate

Corporate Training for Business Growth and Schools