MSFS tops Navigraph 2023 survey results

[GoranM 3,799]

55 minutes ago, threegreen said:

So for the sake of the hobby, I sure hope Meyer does whatever it takes to at least keep XP where it is right now.

X-Plane isn't going anywhere.  Austin and his staff have actually said they get more motivated to keep pushing it forward when they read or hear about the X-Plane bashing.  Not to mention Laminar is a private company owned by 1 person.  Austin Meyer.  MSFS is ultimately owned by shareholders.  And if it goes into the red... 🙄

Anyway, let's not think about that.  I'm sure MSFS will keep going for many more years.

[abrams_tank 7,473]

33 minutes ago, fsiscool said:

The Azure AI detected not masks (areas) of vegetation but every single tree. The total number of 2 trillion trees, which they found on the whole earth, was widely published in early articles about MSFS.

 

15 minutes ago, turbomax said:

not to mention the 4.7 quadrillion blades of grass, they had to cut during the beta because they were found to be too long. they had 500 people working on it.

Wow, 2 trillion trees and 4.7 quadrillion blades of grass, did they mention that in the Blackshark AI and MSFS partnership video I posted before? If it wasn't mentioned there, do you remember the source of this, if it's not too much trouble for the source?

Edited January 2 by abrams_tank

[turbomax 3,307]

Just now, abrams_tank said:

do you remember the source of this, if it's not too much trouble for the source?

the  source was me, myself and I. I estimated them, but gave up at 4.7 quadrillion.

[abrams_tank 7,473]

1 minute ago, turbomax said:

the  source was me, myself and I. I estimated them, but gave up at 4.7 quadrillion.

🤣

[turbomax 3,307]

hI Abrams_tank

don't take any of my posts too serious.

Happy MSFS 2024!

[tonywob 4,770]

48 minutes ago, fsiscool said:

The Azure AI detected not masks (areas) of vegetation but every single tree. The total number of 2 trillion trees, which they found on the whole earth, was widely published in early articles about MSFS.

It's the masks that are in MFS (compiled into the CGL files as rasters)... it would be pretty much impossible to detect individual trees in a forest to any degree of accuracy without additional data (LiDAR, etc), and this shows pretty well in the sim that it's not the case, with lots of false positives or misses (e.g. Detecting crops on fields as trees, etc). I think that statement was just some sort of marketing stunt personally, but it doesn't matter either way, it's still a great achievement and adds a lot to the sim.

Grass is done by shaders and colour, and saying AI detected every single blade of grass is ridiculous. etc.. I think this will be taken to the next extreme in 2024, with more detected, rocks etc (but that's just me guessing).

Edit: I see the grass post was a joke....

 

[abrams_tank 7,473]

7 minutes ago, turbomax said:

hI Abrams_tank

don't take any of my posts too serious.

Happy MSFS 2024!

I misread when you said "they had to cut during the beta," I thought you meant they removed grass from the beta, not that they cut it 🤣

Anyways, have a happy new year to you too 👍

Edit: You got @tonywob too 🤣

Edited January 2 by abrams_tank

[abrams_tank 7,473]

53 minutes ago, fsiscool said:

The Azure AI detected not masks (areas) of vegetation but every single tree. The total number of 2 trillion trees, which they found on the whole earth, was widely published in early articles about MSFS.

I looked for a source on the 2 trillion trees. I saw some articles that mentioned there were 2 trillion trees in the world generated by MSFS, but those articles didn't say how those 2 trillion trees were generated. I did read this source though from the Smithsonian Magazine:

Quote

Just as Asobo scanned individual aircraft, the team used high-resolution satellite imagery, aerial photography, and in many cases detailed photogrammetric data from Bing Maps to re-create the globe. The size of this data set, which is continuously updated and expanded, is now at three petabytes (and growing). Most of the data are vertical (looking straight down), but some are oblique, and all are in two dimensions. In order to present this information in three dimensions and ensure visual continuity, the Flight Simulator team decided to tap the services of Austria-based Blackshark.ai, a company with 50 computer specialists who developed proprietary machine-learning algorithms they fed into the Microsoft Azure cloud computing service. Initially helped along with human input, the Blackshark systems eventually learned to render and color all aspects of the planet, including determining the heights of trees (all 1.5 trillion of them), removing shadows of clouds from aerial imagery when appropriate, and ensuring accurate color. The end result is Earth, a planet rendered to a resolution of at least three feet, with the most populated 300,000 square miles rendered even finer, with a resolution of between three and 10 inches, including 58,000 square miles (mostly cities) rendered in such detail you can practically walk the streets.

So it looks like Blackshark AI was able to determine the heights of the 1.5 trillion trees, according to that article. That's pretty impressive, IMO.

Edited January 2 by abrams_tank

[turbomax 3,307]

the only serious part in my post was indeed a quote during the MSFS beta back in early 2020, when they posted in the beta forum "we cut the grass". in the early beta the grass was too long and touched the wings of low wing airplanes like the Cirrus.

[tonywob 4,770]

3 minutes ago, turbomax said:

the only serious part in my post was indeed a quote during the MSFS beta back in early 2020, when they posted in the beta forum "we cut the grass". in the early beta the grass was too long and touched the wings of low wing airplanes like the Cirrus.

Yep, I remember this. I was one who also reported it, it was crazy, the grass would reach the windows of the plane.

I hope they also hire some lumberjacks as well for MS2024.

[abrams_tank 7,473]

I found a good article that discusses in more detail how Blackshark AI generated the 3D buildings from the aerial and satellite imagery:

Quote

“So it was easy for us to fill positions like a PhD in rooftop reconstruction,” Putz said. “I didn’t even know this existed, but this was exactly what we needed — and we found two of them.

“It’s easy to see why reconstructing a 3D building from a 2D map would be hard. Even figuring out a building’s exact outline isn’t easy.

“What we do basically in Flight Simulator is we look at areas, 2D areas and then finding out footprints of buildings, which is actually a computer vision task,” said Putz. “But if a building is obstructed by a shadow of a tree, we actually need machine learning because then it’s not clear anymore what is part of the building and what is not because of the overlap of the shadow — but then machine learning completes the remaining part of the building. That’s a super simple example.”

While Blackshark was able to rely on some other data, too, including photos, sensor data and existing map data, it has to make a determination about the height of the building and some of its characteristics based on very little information.

The obvious next problem is figuring out the height of a building. If there is existing GIS data, then that problem is easy to solve, but for most areas of the world, that data simply doesn’t exist or isn’t readily available. For those areas, the team takes the 2D image and looks for hints in the image, like shadows. To determine the height of a building based on a shadow, you need the time of day, though, and the Bing Maps images aren’t actually timestamped. For other use cases the company is working on, Blackshark has that and that makes things a lot easier. And that’s where machine learning comes in again.

“Machine learning takes a slightly different road,” noted Putz. “It also looks at the shadow, we think — because it’s a black box, we don’t really know what it’s doing. But also, if you look at a flat rooftop, like a skyscraper versus a shopping mall. Both have mostly flat rooftops, but the rooftop furniture is different on a skyscraper than on a shopping mall. This helps the AI to learn when you label it the right way.”

And then, if the system knows that the average height of a shopping mall in a given area is usually three floors, it can work with that.

One thing Blackshark is very open about is that its system will make mistakes — and if you buy Flight Simulator, you will see that there are obvious mistakes in how some of the buildings are placed. Indeed, Putz told me that he believes one of the hardest challenges in the project was to convince the company’s development partners and Microsoft to let them use this approach.

“You’re talking 1.5 billion buildings. At these numbers, you cannot do traditional Q&A anymore. And the traditional finger-pointing in like a level of Halo or something where you say ‘this pixel is not good, fix it,’ does not really work if you develop on a statistical basis like you do with AI. So it might be that 20% of the buildings are off — and it actually is the case I guess in the Flight Simulator — but there’s no other way to tackle this challenge because outsourcing to hand-model 1.5 billion buildings is, just from a logistical level and also budget level, not doable.”

Over time, that system will also improve, and because Microsoft streams a lot of the data to the game from Azure, users will surely see changes over time.

 

So I guess based on this article, Blackshark AI not only used OSM data to generate the 3D buildings, but they were also looking at aerial or satellite photos, and analyzing the shadows from the buildings, to determine the height of the buildings.

Edited January 2 by abrams_tank

[tonywob 4,770]

2 minutes ago, abrams_tank said:

So I guess based on this article, Blackshark AI not only used OSM data to generate the 3D buildings, but they were also looking at aerial or satellite photos, and analyzing the shadows from the buildings, to determine the height of the buildings.

Yes.. if you want to read more in depth about the various techniques used, one example is here: https://www.sciencedirect.com/science/article/abs/pii/S0924271616301939 there has been a lot of work on this over the years

[flying_carpet 666]

I would like to know how exactly they detected the gazillions of pebbles coming in MSFS 2024 😄 (but maybe that should be discussed in a separate thread, as this one is about MSFS 2020 ...).

[fsiscool 252]

2 hours ago, abrams_tank said:

Wow, 2 trillion trees and 4.7 quadrillion blades of grass, did they mention that in the Blackshark AI and MSFS partnership video I posted before? If it wasn't mentioned there, do you remember the source of this, if it's not too much trouble for the source?

You can google this [msfs "2 trillions" trees] to find many sources. E.g. this one: Microsoft Flight Simulator: The Future of Game Development

 

[Krakin 5,846]

I bet you some people wouldn't laugh if Austin said XP12 was about to get a gazillion pebbles. Laugh on lol.