Claviateur

I think MSFS business model as a PC flight sim will be the end of VFR scenery as theaters or patches...

If as you said, people will be disappointed from the scenery of MSFS, then it will be the most bizarre thing I will have to witness in the history of flight simulator software... I understand that many do not know technically speaking how things are done in the sim engine. Probably they think the world will get mostly photogrammetry when the first version is out or the AI is putting everything where it should be and as it looks in real life... But we must know now that most of the planet will be just like the Sydney screenshot (above) and this is by itself a spectacular thing... It"s a very obvious thing to say that no simulator among those available today, features a default scenery for the whole planet with aerial imagery, details and immersion... Probably many will have to keep their actual simulators installed just to compare for few seconds their area of interest... And I am sure they will then live happy ever after... 🙂

Just came across a new freeware scenery Kodiak Alaska released for the other simulator. Among the few airfields, Larsen Bay is featured with an Orthophoto ground layout as well as most probably an overlay scenery generated with OSM data...

Well yes and no... I noticed in the promotional videos that the Photogrammetry cities LOD is updating and morphing just like it does in Google Earth ( since it"s like a mesh terrain). So it has its obvious advantages when it comes to realism (depending on resolution) but its (LOD refreshing) downside as well... I just hope the global terrain elevation does not refresh like this! So not having a photogrammetry city is not a show stopper at all. The evidence is in this screenshot of Downtown Sydney created procedurally and with footprint data. And it looks superb as an out of the box default scenery...

I suppose if there is a company that should know or have solutions on how AI could be tomorrow's smart photo retoucher is Adobe 🙂

This is a reality we must accept as a compromise in aerial imagery / Orthophoto scenery. Imagery shadows, mix of color hues and seasons as patches of imagery in the same region as well as the imagery static snow on the mountains all year long will be there... But... I used to think this would bother me a lot before I started generating Orthophoto scenery for the other simulator. Honestly, when I started flying over that generated Ortho scenery, all these were not a show stopper at all... Once in the simulator's engine, rendered with vegetation and all the 3D infrastructure and the details, all these bothering stuff will not be visible or noticeable... At least this is what I experienced myself... To me Ortho scenery with all these baked elements is still way more realistic than any generic texture based scenery in the actual simulators... If it has the 3D elements over it. Indeed 🙂

I remember when OrbX released their FTX Global scenery for FSX, users reported like a million flaw in the places they visit on this planet, all sort of glitches here and there from the data etc.I assume the same thing will happen with MSFS. But I hope user friendly tools will be available and users will learn the DIY scenery improvement for their area of interest. Otherwise poor Asobo 😕

I want this in MSFS 😛

Now a new forum is required and it could be called The "Alpha / Beta Market" 😁

This is a news report about Asobo's Plague Tale video game winning many Awards... At the end of it, a small mention of MSFS... It's in French 😛

You are right, the BlackShark eventual partnership made me wonder why MS wants an external studio to manage its Azure AI imagery scanning and data building for a virtual world generation. I am wondering if they want to take the whole thing beyond MSFS. I do not know what agreements they have with Asobo but who knows, the world engine could be a long term project for more than just our dear simulator. Or is it that internally they are not able to manage the data augmentation demands Asobo requires for their features. Interesting story to follow for sure.

I was most probably mistaken about vegetation zones vs point based scanning 🙂 Now I don't want to suggest anything here but I will be very very very surprised and impressed if the AI brings height and types! My suppositions that could be again wrong, is that these are based on Asobo's procedural script rules and or from another world vegetation census but, again, I could be wrong... Again...

🧐

Yes. In fact the trees evolved from evergreen only to be more or less 3 types (for now)?

Did I miss this screenshot in yesterday's update?

Now that I open my eyes wider and look more carefully at the screenshots with vegetation zones, as Shack95 said, Asobo are using vegetation points and not areas (polygons). To my knowledge, vegetation polygons can`t reproduce the result we see here... It`s more precise thus it is amazingly generated as it seems, from vegetation points from the AI scanning. I was mistaken and if this is the case, then this is so impressive indeed.

Now you're inspiring those who were not invited to the Alpha, to post "I don't care anymore" on the forum here 😁

The AI today is capable of doing amazing things but the question is always per project / budget and pragmatism I suppose. For MSFS, as Shack95 clarified, the Azure AI is building a database by fetching data from scanning and putting it on the desk of Asobo to work with. The clouds for example are detected and tagged and Asobo deal with the retouch in some way or another (manual or Photoshop scripting who knows). So as I initially thought for MSFS, the Azure AI is a data fetcher and this is a great thing already. And I suppose this makes full sense because creating a procedural world from traditional scripting (or node base scripts as in modern engines) using data from any source including AI scanning, is less complex in my opinion than having an AI that builds a scenery visually, retouch imagery etc. For now at least and the upcoming few years, I think it's a cheaper and more logical for a video game (MSFS or any other MS game project) to continue with traditional development methods while using machine learning for data augmentation as Asobo mentioned... But will the AI one day be a video game creator (code and visual)... No doubt!

Thanks for the infos, I missed the detail about individual tree tagging. This is impressive and huuuuuuge! I wonder how they handle this data or, perhaps, as I mentioned previously, they could group them into zones. But point based database for the world forests is a huuuuuge data to work with and I can't wait to see how it was translated into 3D assets.

Let's say I put some AI at work to scan the Bing aerial image of Larsen Bay, I do not know what zoom level the MS Azure AI uses but let's say the one I grabbed in the screenshot below... I think and based on my observation of the screenshot the vegetation zone is more or less what I draw manually with red lines in this small portion I don't think the detection is done one green patch at a time otherwise the data will be extreeeemly heavy for a small region. Edit: If it's detected one green patch at a time, I suppose grouping is done for processing matters and data heaviness. So I presume the empty patches detected, and as seen in the screenshot, are no smaller than this (minus roads)

You might be right and the data fetched by the AI might be very sophisticated as we speak but if this is the case, then the reason why we see what we see in this or any screenshot, could be caused by at least one of the following reasons, based on my knowledge with how this global scenery is done (from any data source) : 1- As I suggested above, the types of vegetation could be (still) limited. Thus no matter how detailed the AI scanning result or any data source will be, if this is not matched by equivalent amount of variated visual assets, the result will remain rough and approximate. And / Or 2- The scripting to generate this procedural world from the fancy data does not have (yet) sophisticated rules or sub-conditions and sub-filters to go deep into small regions on this planet. Edit: The rules and filters could go down to a specific vegetation zone(s). i.e: using all its tags or making it one type... Or grouping small zones together for the sake of performance... So data alone, as fancy as it could be, would never by itself give a result because it must be translated into a visual representation using scripting + visual assets... So I could be mistaken but in MSFS the AI does not generate visual scenery. Asobo do from the available data with scripting they type themselves for the engine to understand and thus to render visual elements...

Absolutely... This global scenery generation process is impossible to control or test on a granular level unless the super detailed data is met with super detailed assets library and super detailed scripting conditions... And still... As you said, me too, as a virtual tourist in this virtual world, I will never know that this pretty landscape is not a 100% match of the real thing because I will never check every step I make in the sim...Well except for my area of interest for which I will create custom stuff... Hopefully...

In Open Street Map (OSM) the vegetation zone added manually by a human contributor, is defined as forest. https://www.openstreetmap.org/query?lat=57.5364&lon=-153.9666#map=15/57.5341/-153.9692 Now I see a very similar zone filled with evergreen in the screenshot of MSFS. I am not saying Asobo used OSM data but the covered zone is roughly the same.

The AI does not add, it scans aerial imagery and determines the vegetation zones (polygons) like the footprints. Depending on the quality of the aerial imagery the zone is precise or rough.

The question is not learning, the question is how many types are available for now in the vegetation library... If let's say you have 3 tree types in your vegetation library for the whole world and your real world database or AI scanning or OSM or any souce features 100 types of trees for the world, you will unfortunately have to divide the 100 types into 3 because your visual assets only feature 3... This an example, of course...