A lot of people still do not consider Pokemon Go “real” AR. At launch, the AR features were extremely limited and in most cases the camera would simply plaster an animated Pokemon over the top of your real world environment which resulted in the creature seemingly standing inside of objects and walking across things it shouldn’t be able to walk across. In short, it lacked real depth and environmental awareness.
Recently Niantic acquired a London-based technology company called Matrix Mill. By using machine learning and neural networks, they’ve more or less solved the issue. Watch the results in the video below:
“At Niantic, we frequently talk about how in order to augment reality, you need to be able to understand it,” says the Pokemon Go developer in a blog post. “The Matrix Mill team has come up with novel ideas that push the boundaries of what machines can process, thinking around occlusions, and seeing the world closer to the way human eyes can. As a result of this hard work, AR experiences can feel more natural to the eye, which is a goal we have squarely in our sights.”
This sort of stuff is possible today using expensive cameras and advanced equipment, but everyday smartphones with single cameras on them like the Google Pixel 2 or the latest line of iPhones couldn’t achieve this effect. Now, with Matrix Mill’s help, Niantic can deliver much more convincing AR.
By interpreting and then re-sending the data back to devices, the depth effect in AR works seamlessly and characters like Pikachu and Eevee can be occluded behind objects in the world or even people walking by in real-time.
We don’t really know the timeline for when this sort of functionality will officially roll out to users, but it’s exciting nevertheless, particularly as the company preps for the eventual launch of their next AR game, Harry Potter: Wizards Unite.
Let us know what you think of this occlusion demo down in the comments below!