Intel Uses Machine Learning To Make GTA’s Graphics Look Scarily Photorealistic

GTA 5’s graphics are decent and it looks good considering how old the game is, but we would never ever call it realistic-looking, but that might not be a bad thing. That being said, if you’ve ever wondered what a game like GTA could look like had it been developed with photorealistic graphics in mind, then you’re in luck.

This is because thanks to researchers at Intel Labs, they have decided to try and apply machine learning techniques to rendered footage from a console game, like GTA, and make it photorealistic. The end results can be seen in the video above, and we can tell you that it looks pretty damn real.

In fact, if we didn’t know that the graphics were enhanced by machine learning, we would have assumed that it was video footage taken in real life. The cars look way more realistic, the roads also look a lot more real, and the vegetation also looks a lot more believable. This is thanks to the use of datasets that the researchers fed to the neural network, allowing it to render images based on real life over that of computer generated ones.

While we will probably still need powerful hardware to help render graphics in the future, the work of the researchers seem to suggest that software could also play a key role in helping deliver enhanced graphics in the future.

Filed in Gaming >General. Read more about AI (Artificial Intelligence), GTA (Grand Theft Auto) and Intel. Source: gizmodo