X-Plane 10 HD Mesh v2 - PREVIEW

[alpilotx]

Here are some first PREVIEW screenshots (mostly from the European Alps) of my upcoming new X-Plane 10 HD Scenery Mesh v2!

https://picasaweb.google.com/101666907909842492197/XPlane10HDMeshV2PREVIEW

(it seems Picasa has somehow blown up the saturation of the images ... don't know why)

 

A short list of the improvements:

• increased mesh density
• improved forest representation (in Europe, reprocessed/improved raw forest landclass data)
• all new OSM data (streets, railways, power lines, water)
• linear OSM rivers included (Tag:waterway=river)
• some - minor - landclass tweaking (like strip mines or riverbeds etc.).

Planned coverage:

• Europe (this time, entire Europe)
• USA (complete)
• Alaska
• possibly parts of Canada (like the West)
• ... etc. (depends on the data size ... but this time the single DSFs will be even larger, so ...)

Release? Hopefully this summer (most of the things are already in place, but there are a few things - partly not in my hand - which need to happen before).

 

[sqrt(-1)]

Wow!  Just WOW!

[chris k]

Andras,

 

That looks TREMENDOUSLY GOOD.

 

My hats off to you and the AutoGen Team....!

 

- CK.

[JohnMAXX]

Andras...

 

This is what we need, Alberts textures with this mesh looks perfect 

[cabnz1]

Wow, great work

[Cameron]

Amazing work! Thanks for showing us some early work!

[LuisAngel]

Thanks!!

[pryoski]

Looking forward to seeing this when I get home! Cheers

[Airbus]

Incredible work!

[pryoski]

Holey smoley! That is incredible!

Edited June 26, 2013 by Kris Pryo

[alpilotx]

UPDATE: I have just re-uploaded all pictures after I finally found the - well - hidden setting in Google, which over saturated all of them (why did they set this as default at all ... argh). Now everything looks normal ....

[Colin S]

If you do Canada, I will officially declare your the awesomest overlord of X-Plane that ever walked the face of this earth. HOLY CRAP THOSE LOOK AWESOME. Looking forward to... well... EVERYTHING............ omigod.

[Tom Knudsen]

One thing i am sure of, i really need to do something about those horrible xp mountain textures.

On another note, hd mesh looks good, what resolution are they at

Sent from my iPad using Tapatalk - now Free

[alpilotx]

Maybe you wait for the texture updates in 10.25 ... and horrible is extremely relative (be very careful in your wording), especially as we are talking about - kind of - artistic interpretation of real life phenomena (which always reflects the taste of the artist too). And I can assure you, that neither is it something done "overnight" ... (I know how much work went into them over the years) ...

 

And maybe you have already read, that there is no such thing as "mesh resolution" in X-Plane when we are talking about an irregular mesh. Maybe this old screenshot from my HD Mesh 1 download page helps you to get a feeling of how / what changes (and HD Mesh Scenery v2 will even have a slightly denser mesh than v1):

 

 

[AtomicFrawg]

One thing i am sure of, i really need to do something about those horrible xp mountain textures.

On another note, hd mesh looks good, what resolution are they at

Sent from my iPad using Tapatalk - now Free

 

LOL These textures are 1000 times better than the stock textures of FSX....... Oh and these are free...

[pryoski]

1000 times better

 

Actually 10,000 times better ... 10,000 has historic precedence to mean 'much much'  or 'lots and lots' better ... that number where things coalesce, crystallize and become congruent ... what I'm actually saying is that I totally agree with your post!

[Tom Knudsen]

One thing i am sure of, i really need to do something about those horrible xp mountain textures.

On another note, hd mesh looks good, what resolution are they at

Sent from my iPad using Tapatalk - now Free

LOL These textures are 1000 times better than the stock textures of FSX....... Oh and these are free...

Who did compare it to default FSX?

I will for sure see what 10.25 looks like, hopefully textures are not shown in this thread

Was not aware that Xplane 10 did not use SRTM DEM data for its elevation mesh, cause if it did i bet your HD Mesh is about 3arc or about 90m DEM for LOD ( Level of detail) What kind of mesh typology X-Plane uses i do not know, but I do tend to use 10m DEM data for a Norwegian mesh when Meshtool finally comes out for XP10. Untill then lets hope 10.25 will improve

Sent from my iPad using Tapatalk - now Free

[alpilotx]

Tom, I have the feeling that you are mixing a LOT of things here ...

First you talk about textures, then about the mesh, the SRTM ... and I would like to know, which of them do you think affects what???

 

So, first: I did nowhere say, that we do not use SRTM data. Of course we use elevation data just like SRTM (earlier we used it directly, now indirectly) or where else would we/I get a planet wide elevation data coverage in a halfway consistent form (you always have to keep in mind, that most of my work is not just local, but has to work in very big regions - possibly on planetary scale)? For HD Mesh Scenery v2 (and all other Laminar work) it is now the planet wide coverage from viewfinderpanoramas.org (you might check that out - it is a mixture of SRTM and many other sources ... usually the best data available for a given region). And yes, that raw data is at 90m resolution.

 

My previous post was about how the mesh looks like at all ... its an irregular triangle mesh network (effectively not even a contiguous mesh, but its comprised of myriads of overlapping little triangle mesh patches). Its irregularity depends on the structure of the underlying elevation data (the raw data which its derived from) ... where homogeneous landscape (with little elevation changes) gets fewer, large triangles, while a landscape (usually mountains) with many changes gets more, smaller triangles. 

 

This triangle mesh is - since XP10 - saved only in its 2D structure (saves some space etc.)! And additionally, each DSF file contains the elevation data in its raw, raster form. Then when the scenery is loaded, XP10 "puts" the 2D triangle mesh over the raster elevation data (like you would put a thin veil over a sand castle on the beach) ... which gives the triangle mesh its final, 3D structure.

 

Its important to know, that effectively the "resolution" of the irregular triangle mesh (as it is in its 2D form in the DSF file) limits how detailed you will see it in 3D (even if you would put more detailed raster data under it) ... BUT there is one possible future remedy for this (it has been planned for a long time, and don't ask me when Laminar will implement it - but its planned): tessellation ... Which would dynamically break up the triangle mesh in much smaller pieces and thus would make it possible to much better "mimic" raw elevation data (without the need to have extremely high res tirangle mesh data to be stored in the DSF file) ... Aaah, and before I forget it: there is already a funny little tool fo X-Plane, which allows you to manually increase the triangle mesh resolution (which then gets stored in the DSF file): http://forums.x-plane.org/index.php?app=downloads&showfile=16417

 

You can read about the mesh and the underlying DEM data, and tessellation in Ben Supniks blog too:

http://developer.x-plane.com/2011/08/dsf-gets-raster-data/

http://developer.x-plane.com/2013/04/dsf-mesh-formats-v9-vs-v10/

 

(or just search it for "tessellation")

 

And finally:

X-Plane 10.25 will ONLY improve the artwork. This means especially - and most importantly - improve a lot of the existing textures (the "pictures" which are painted on the existing 3D mesh!), bring lots of new textures and possibly some new autogen artwork. The new textures will also include some experimental new urban textures (which will be visible in some regions - depending on climate). X-Plane 10.25 WILL NOT change the scenery itself nor how it is rendered.

 

New scenery, with higher resolution mesh and based on new OSM data will be available vial my HD Mesh Scenery v2 ...

[AtomicFrawg]

Who did compare it to default FSX?

Sent from my iPad using Tapatalk - now Free

 

I did...

[Tom Knudsen]

You can read about the mesh and the underlying DEM data, and tessellation in Ben Supniks blog too:

http://developer.x-p...ts-raster-data/

http://developer.x-p...mats-v9-vs-v10/

 

First off, see the latter link, Ben is answering my question in a previous blogpost, second most of what you are explaining are more or less quoted from that post.

 

 

 

For HD Mesh Scenery v2 (and all other Laminar work) it is now the planet wide coverage from viewfinderpanoramas.org (you might check that out - it is a mixture of SRTM and many other sources ... usually the best data available for a given region). And yes, that raw data is at 90m resolution.

 

Why is that better than the original raw shuttel mission data from NASA?

http://www2.jpl.nasa.gov/srtm/

 

 

 

 

First you talk about textures, then about the mesh, the SRTM ... and I would like to know, which of them do you think affects what???

 

In case you already havn't noticed, i am talking about mesh not texture

 

From Bens post

 

 

Mesh: In X-Plane 8 and 9, the terrain mesh is stored as a set of triangles in 3 dimensions; each corner of the triangle has a latitude, longitude, and altitude.  The shape of the mesh comes from the location of those triangles and the heights of each corner.

 

Mesh + DEM: X-Plane 10 can also handle a new extended DSF with raster (array) data.  In this mode, the mesh contains triangles (just like it did) but they contain only latitude and longitude.  The elevation for the entire DSF tile is stored in a 2-dimensional array of elevations (a raster DEM).  When X-Plane 10 loads this format, it reads the height for each triangle out of the array of elevations to “rebuild” the 3-d triangles at load time.

 

X-Plane 9 supports only the original “mesh” DSFs.

X-Plane 10 supports both the original mesh DSFs and the new Mesh + DEM DSFs.

 

First som facts

 

1. DEM data for the entire world right now is 90meter as it has been since v9 right?

2. We have only Meshtool for v9 i.e it kan only support mesh DSFs.

 

So my questions are..

 

If you are able to modify the 2d array as shown on the picture above, would not this be modifying just the lat and long quality and not the exactly elevation (would be with respect why i did not notice any difference in elevation quality with HD Mesh v2 vs orgininal mesh)

 

If you are able beyond this to alter the 2D DEM array to even calculate the elevations (raster DEM), would it not be better elevation quality (i.e LOD - Level of detail) with better DEM data such as i.e 75 meter for the whole world?

 

 

 

So Textures

 

As you said, this is just pictures placed on top of the 2D raster layer as textures normally are, even with bump maps or speculars.. Anyway

The clue must be to individually fit this to each triangle right? But that just if I am correct only affects placement or purhaps distortions???

What I am curious about is what textures can be modified in order to change the picture quality on a global scale?

 

I did this just as an experiment for the states, but dropped it for sake of some OSM scenery while a go, but purhaps I need to retry that experiment.

 

 

X-Plane has about 100 land-uses for the whole planet. (Desert, forest, farm, etc.)

X-Plane applies terrain textures to those land-use types.

Just open the "landuse.txt" to see what textures are used by what land-use types.

 

Could not be that difficult, just time consuming which again should be able to be automated

 

 

I am curious not critising for fun, hope you take my questions as productive and not an personal attack he he..

 

PS..

Keep up the good work nevertheless.

 

Here is btw a master theses on the subject of tessellation

http://dice.se/publications/adaptive-hardware-accelerated-terrain-tessellation/

Edited October 8, 2013 by Tom Knudsen

[alpilotx]

Tom, the most important thing which you maybe don't realize here is: I am - even if only as a freelancer - working "inside" Laminar since many years (you find my name in X-Plane). I have done most of the data work (preparations, rule systems etc.) for the default Global Scenery and my HD Mesh works are derivatives ... and now even many improvements over that work. And no, I do not work with meshtool, but with the Laminar internal scenery generator called "RenderFarm" (don't ask me why this is its name ... it was not me who named it ) ... So.... I do not change an existing Global Scenery mesh, neither do I change "manually" (or in post processing) its landclass use. But instead I work with RenderFarm, which generates genuine, new DSF scenery tiles from the wast amount of data I feed it.

 

The areas where I do my improvements are (and this list is for sure far from complete):

• Improve the input landclass data (there are many different sources for different regions of the planet ... all with their many pros and contras ... and non of them easily adaptable to the DSF creation process).
• Tune / change the complex rule system which tells RenderFarm how to "translate" the raw data into scenery (especially how and which terrain types to apply where, depending on slope, climate and landclass information etc. etc.) ...
• Work on vector data (import better / new sources ... at the moment I work on fresh OSM data import)
• Sometimes even do a little hacking in RenderFarm itself (its written in C++ and usually its Ben Supnik who hacks it). ... Like I did with a new algorithm which makes the new, line river based small rivers look halfway sane in mountainous areas (its always a mess, when the line vectors don't match the elevation mesh exactly ... then rivers run up and down over hills etc.)

So ... you see, this is all not easy work (but quite interesting - otherwise I wouldn't do it) and neither is it the way - at least thats my feeling when I read your comments - you think its done

 

Now to some of your other questions:

 

Yes, indeed, the denser elevation mesh - very simply put - only changes the 2D structure of the triangle mesh (it makes smaller triangles) .... But then you didn't understand my second description (might easily be my fault - I am not a native English speaker). Namely, how X-Plane itself combines at runtime (well, when it loads the DSF) the 2D triangle mesh with an added elevation raster layer (effectively the 90m DEM data) .... and turns the 2D triangle mesh into a 3D triangle mesh (in X-Plane 9 this was different ... there the mesh was saved directly in 3D). And thats why I told, that effectively the "resolution" of the irregular mesh is the "limiting factor" when it comes to representing elevation detail. Because X-Plane - at least until now - can only make it as detailed as the triangle sizes (in the DSF) allows it (so, if you would put in 30m or 10m raster DEM into the DSF, that would only waste space ... the triangle mesh couldn't use it). And this is the point, where tessellation can help in the future ...

 

One more very important aspect of the triangle mesh is, that neither is it contiguous (its not one big triangle mesh) .... but rather comprised of myriads of small patches. Usually each patch can have one terrain type assigned to it ... and thats how you get all the different textures in the final look (and yes, its the RenderFram which makes those patches have the right size to  - indirectly - represent landclass etc. as closely as possible). On  my website you can see a screenshot which gives you a feeling how many different patches a DSF tile can have (its on the site of a tiny script I wrote, which allows you to translate DSFs in a form, which the tool QGIS can visualize - this tool can help you to see which terrain type is applied where in a given DSF):

http://www.alpilotx.net/downloads/tools-scripts/

 

Second: about viewfinderpanorams.org. Well, there are many reasons to be better than the original SRTM data. Like for example many many holes and data errors in the original SRTM90 data. The limitation to only have data up to 60N (most of Norway / Sweden / Alaska etc. etc. would not be possible with only SRTM90). What viewfinderpanoramas.org (its effectively one guy - I had some conversations with him over time) does is, to get elevation data from many many different sources (if possible better sources than SRTM90) and merge it together to have one final dataset which has an overall good quality (in many areas far better than SRTM90 .... believe me or not ... you can have big quality differences even at the same resolution!), no holes, and a complete coverage of the planet ... which takes the burden off my shoulders to do similar things on my own (which I did for the first Global Scenery for XP10 ... and now I am happy that I will not need to do that anymore and still have better data).

 

Then meshtool .... well, as we / I am not working with meshtool since a long time (well, never used it to be honest), and XP10 has a completely different terrain system (vastly improved since XP9 ... and meshtool is only really compatible with XP9 style terrain) you can forget about the "100" land uses ... if you look in the folder "Resources/default scenery/1000 world terrain/terrain10/" you will see that there are already over 3500 TER files (effectively each representing a possible, different terrain type .... usually a combination of slope / climate / landclass ... and a few other possible factors). Of course, you will never have a single DSF referencing all of them at once (always only a subset).

 

And the textures   ... neither are they just simply painted on the triangles. In XP10 there are many ways - via so called shaders - to apply the textures in many different ways. Like for example rotate them in the direction of the slope (used extensively in mountains), or combine two textures to make a more random look, or apply the from the "side" (rather than from top) when using them on very steep terrain like the cliff (its even called the "cliff shader") etc. etc. ... So, you see, even this part is far from trivial. And our texture artist is not just creating the textures, but also invests lots of time to parametrize all those terrain types to use the right shaders with the right parameters (there are quite a few parameters which can be tweaked) to get it all look as good as possible ... And no, this is not done by hand editing each TER file, but via big tables (we use OpenOfffice for that) which makes it halfway comfortable for him to do the parametrization ...

 

By the way, I also wrote about some of these things in a very old interview, long time ago for xsimreviews (there are some old infos about the DSF, and scenery generation process ... but still, lots of that info - even if not all - still apply):

http://xsimreviews.com/2011/12/10/developer-interview-andras-fabian-mr-x-terrain/

[Tom Knudsen]

First off, let me say I am deeply grateful that you share your knowledge!

Second, not a english native my self so excuse any missunderstandings!

 

 

The areas where I do my improvements are (and this list is for sure far from complete):

• Improve the input landclass data (there are many different sources for different regions of the planet ... all with their many pros and contras ... and non of them easily adaptable to the DSF creation process).
• Tune / change the complex rule system which tells RenderFarm how to "translate" the raw data into scenery (especially how and which terrain types to apply where, depending on slope, climate and landclass information etc. etc.) ...
• Work on vector data (import better / new sources ... at the moment I work on fresh OSM data import)
• Sometimes even do a little hacking in RenderFarm itself (its written in C++ and usually its Ben Supnik who hacks it). ... Like I did with a new algorithm which makes the new, line river based small rivers look halfway sane in mountainous areas (its always a mess, when the line vectors don't match the elevation mesh exactly ... then rivers run up and down over hills etc.)

 

Seems like more fun than work, but then I do not work with it so to me this is just fancy.

 

I tend to get annoyed with things I do not understand and furthermore go to extended distances to figure things out. In some cases it will in fact be to annoy the heck out of people like you

and Ben with question about things I want to figure out or do not understand.

 

So how come X-Plane only have 90 meter elevation data where simulatators like FSX and Prepare3D claim down to 10 meters, 38 by default

 

 

Second: about viewfinderpanorams.org. Well, there are many reasons to be better than the original SRTM data. Like for example many many holes and data errors in the original SRTM90 data. The limitation to only have data up to 60N (most of Norway / Sweden / Alaska etc. etc. would not be possible with only SRTM90). What viewfinderpanoramas.org (its effectively one guy - I had some conversations with him over time) does is, to get elevation data from many many different sources (if possible better sources than SRTM90) and merge it together to have one final dataset which has an overall good quality (in many areas far better than SRTM90 .... believe me or not ... you can have big quality differences even at the same resolution!), no holes, and a complete coverage of the planet ... which takes the burden off my shoulders to do similar things on my own (which I did for the first Global Scenery for XP10 ... and now I am happy that I will not need to do that anymore and still have better data).

 

Well there are many tools out there that fixes SRTM data (patching holes and filling in like above N60) i.e SRTMFill http://3dnature.com/srtmfill.html and SRTMVoid Killer http://www.dgadv.com/dgvk/

But thats not the point!

 

 

What i am not sure about is if it even exists better data for the world than SRTM90 3arc collected by Endeavour back in 2000 (except SRTM30 for the US that is).

Or if they did in fact to a new dataset collection on the last shuttlemission.

 

 

I know one thing is to want someting better, but my frustration is that I do not know if it is even possible.

As you said, purhaps tessellation is the only way out.

 

 

PS..

 

How the heck can Outerra do this type of mesh

http://www.simflight.com/wp-content/uploads/2012/03/k298.jpg

 

 

where FSX can do this

http://www.hardwarecanucks.com/forum/attachments/gaming/1978d1209209108-fsx-add-sandes_sion2.jpg

 

 

but XP only this

https://lh3.googleusercontent.com/-aj3OyrOON3k/UjTndfsL_EI/AAAAAAAADGw/0w5_Z7rZp3A/w1597-h832-no/FA-22A_35.jpg

 

 

NB!

 

Could you do something about the mountain texture only? Looks like someone has sneezed on it with their now full of cocain LOL..

https://lh6.googleusercontent.com/-VS52wJqGyrQ/UjTnfb6e3pI/AAAAAAAADHU/W6xlVI27TTI/w1597-h832-no/FA-22A_41.jpg

Edited October 8, 2013 by Tom Knudsen

[Rhard]

Well there are many tools out there that fixes SRTM data (patching holes and filling in like above N60) i.e SRTMFill http://3dnature.com/srtmfill.html and SRTMVoid Killer http://www.dgadv.com/dgvk/

But thats not the point!

 

 

What i am not sure about is if it even exists better data for the world than SRTM90 3arc collected by Endeavour back in 2000 (except SRTM30 for the US that is).

Or if they did in fact to a new dataset collection on the last shuttlemission.

 

 

I know one thing is to want someting better, but my frustration is that I do not know if it is even possible.

As you said, purhaps tessellation is the only way out.

Hi Tom,

 

all this tools just interpolate the elevation data. Most of this data are based on SRTM4 data sources. So even if you use 30m or 10m raster layer it will give only more smooth surface but not much details. Outerra creates this details based on fractal algorithm which looks good but is not really true.

Another things is if you find real HI resolution elevation data. I think there are some and Alpilotx probably can tell more about it. (NZ for example).

 

Here is my question to Alpilotx:

1) can RF use HI resolution DEM to create raster layer for XP10 DSF?

2) Is it possible to combine two or more elevation data to create one DSF? Or the raster layer is a single instance with fixed resolution for entire DSF?

3) Is it possible to add some data to the RF, that it will add XP9 style mesh only for specified area (i.e. airport fields) and rest will remain as XP10 2D triangulation. (I know that now some points in DSF still have an elevation value - i.e. river and lakes borders).

 

Sorry for asking you, but Ben is quite busy and I don't want do disturb him. I already compiled DSFTool and MeshTool and have some ideas how to play with code and add some features. Now I want to try compile RF again.

[Tom Knudsen]

So what if I had access to 10meter DEM data (i do but not sure if it can be publicly used yet), could it be used in XP?

PS

I think Austin should buy Outerra and merge them for sure

Sent from my iPad using Tapatalk - now Free

[Rhard]

PS

I think Austin should buy Outerra and merge them for sure

But what should be with all Ben's work in this case?? For sure he will never do that.

And X-Plane engine is good as well, but not really adjusted.. Terrain textures (texture itself and way they are combined in pattern) by default is what I can't imagine even in my worst dream. Sorry.

Only Alpilotx staff gives me a hope for the future of X-Plane.

Edited October 9, 2013 by Rhard