Setting up a Gitlab Australian scenery group?

[frtps]

Hi all,

I'm thinking of setting up an Australian scenery group on Gitlab. It would act as a hub for Australian scenery development, with the idea that mature work would be fed back into flightgear, much the way that aircraft development seems to run. I'm aware that this is not the first attempt to do this for Australia in the long history of flightgear, but perhaps the open git environment and feedback into flightgear will make it worthwhile. It may even help to develop a model for other areas of the world.

Things I would plan to cover:

• Division of Australia into scenery regions
• Collection of images/links to images for texture creation
• Creation of recipes for automated scenery generation based on Australian data sources + OSM
• Creation of textures for terrain, osm2city objects
• Collection of generic objects (ie no American barns in the Australian outback)

Any thoughts on this? Is there already something similar somewhere?

And because no scenery post is complete without a gratuitous screenshot, here is a shot showing two textures I've been working on, one for cliffs and one for Australian bush.



Cliffs in the Blue Mountains west of Sydney. OSM + SRTM-3 data.

[wlbragg]

no American barns in the Australian outback

Oh, come on, there has to be at least one crazy American that built a barn in the Outback!

That scenery looks really good. You seem to have captured the feel of a totally different region and climate. Well done.

[frtps]

That scenery looks really good. You seem to have captured the feel of a totally different region and climate. Well done.

Thanks for the encouragement. Whether that totally different region exists in real life is another question! The bush seems too gray and the cliffs are washed out compared to the source texture (and real life). The vegetation source texture came from an openaerialmap.org ortho photo of bush in Western Australia so in theory the colours should be closeish, I'm probably going to have to fiddle around in gimp.

[xDraconian]

That's a good looking start frtps.

I like your approach toward organizing a large scale scenery project. The scenery contributors make up a small group, and there are already a couple of regional projects in the works, so it may take some time for others to join in.

[wkitty42]

Cliffs in the Blue Mountains west of Sydney. OSM + SRTM-3 data.

dude! nice pic!

[vnts]

Collection of images/links to images for texture creation
Creation of textures for terrain, osm2city objects
..
Cliffs in the Blue Mountains west of Sydney. OSM + SRTM-3 data.

I recall there was WiP effects & textures for desert regions: (thread. May be something that is useful to modify / build on, for desert & more arid regions of the continent.

Looks great, btw (the blueness is from aerosol released by trees IIRC)

OSM cliffs work really well too . Does the same way of texture mapping work for OSM beach data or even river banks(?) , conceptually?

Kind regards

[frtps]

I recall there was WiP effects & textures for desert regions: (thread. May be something that is useful to modify / build on, for desert & more arid regions of the continent.

Looks great, btw (the blueness is from aerosol released by trees IIRC)

OSM cliffs work really well too . Does the same way of texture mapping work for OSM beach data or even river banks(?) , conceptually?

Thanks for the desert pointer, I hadn't heard of that work. Yes, I had heard that the blueness is from the eucalyptus. I'll have to get into the effects to see if the colour of the haze is configurable.

OSM beach data is just polygons, that are textured based on global latitude and longitude, e.g. ocean-side is not known at texturing time. Rivers are interesting because they are based on line data, and the direction of flow for OSM-sourced data corresponds to the direction of increase of one texture coordinate. However, this coordinate starts fresh from each rectangle that makes up the river so you wouldn't overlay a texture image like I have done with the cliffs because the starting point would change and not match up smoothly. The fun thing is that when a river goes over a cliff there is a single rectangle from top to bottom, so it could be textured as a waterfall using simple shader tricks if a river shader checks the normal to the rectangle (ok for the pedants the normal to the triangle). As a stretch project, the weather generating routines could detect large waterfalls and provide some local mist for rain drops and rainbows. And unicorns too please.

This is sort of off topic but the other experiment I thought of trying is defining a "road verge" and "river bank" landcover type. When you decode the road/river data, you could output both a broad "verge" strip of polygons and a narrower "road/river" strip of polygons. If you make the road/river higher priority than the verge type, the road/river sits on top of the verge, and the intersections should come out ok (i.e. no strips of trees running across roads). This is relevant to Australia in that many of our country roads and creeks are only detectable from above as lines of trees in a cleared landscape, the road or river itself is often partially or completely hidden.

[vnts]

However, this coordinate starts fresh from each rectangle that makes up the river so you wouldn't overlay a texture image like I have done with the cliffs because the starting point would change and not match up smoothly

I was just wondering about a coordinate across the river instead of along it. To let next gen scenery methods run: shallows, banks, waves (??).

OSM beach data is just polygons, that are textured based on global latitude and longitude, e.g. ocean-side is not known at texturing time.

I guess checking elevation data for differences is too small to be reliable, or not possible.

Yes, I had heard that the blueness is from the eucalyptus. I'll have to get into the effects to see if the colour of the haze is configurable.

Just mentioned it because of discussion on colour & interesting phenomenon. Assumed regional haze definitions wasn't currently doable..

If shaders supported low haze colour already (?) they need to be told by uniforms (or vertex attribute data). Some blue haze have to apply, and smoothly transition, to adjacent unforested terrain, islands of different terrain like rock, and objects like roads or buildings (?). That may be difficult through the regional effects system for terrain only. If haze colour is possible at all, it may only work via landclass sampling by the weather system. Not sure..

(I have only heard about the cause of blueness: secreted oil combining with dust, and water vapour somehow. Searching, google appears to think the strong blueness with certain tree species is strongly contributed to aerosols formed by released hydrocabon oils (terpenes) which are compounds of isoprene 1 2,3. Areas that contain isoprene emitters aren't only restricted to Eucalyptus forests it seems. Oaks, ferns, mosses & poplars also emit lots of isoprenes. Isoprenes are 1/3 of total hydrocarbons released. Amount is as high as Methane. The Blue Ridge Mountains 3 have isoprene emitters (Oaks), and blue haze caused by forests occurs elsewhere too in France, Italy etc. e.g. 4. Unexpected )

The bush seems too gray and the cliffs are washed out compared to the source texture (and real life).

Bush there is light ground and darker vegetation layers of different shades: Google maps 1. Low res photos can mean colours of new & old leaves, different species leaves, ground colour, trunk colours, bush/plant/grass colours, get mixed. AIUI Sensors on satellites and whatever, use bands that are optimised for commonly interesting information, not perceptually correct colour reproduction. There is a lot of post processing, so other aerial sources may give better hints. AIUI procedural texture mixing & noise texturing can add detail at different scales 1,2. Small plants & detail on cliffs too.

The fun thing is that when a river goes over a cliff there is a single rectangle from top to bottom, so it could be textured as a waterfall using simple shader tricks if a river shader checks the normal to the rectangle (ok for the pedants the normal to the triangle). As a stretch project, the weather generating routines could detect large waterfalls and provide some local mist for rain drops and rainbows. And unicorns too please

Maybe (?) it can be done with existing tech? A small attachment landclass area can allow a 3d waterfall particle emitter to be automatically placed by default? Texture coords & vertex attributes etc. would have to provide orientation, height, guess at flow velocity, cross-section. Don't know about mist, rainbows. Existing waterfall particles may also do mist(?). Weather tech may be able to do rainbows if a rain (droplet) volume is defined when detecting a waterfall attachment landclass.

Not sure, Thorsten or someone familiar with FG systems may be able to tell

Unicorns just might be much easier A random scenery object. IIRC there was a horse model

This is sort of off topic but the other experiment I thought of trying is defining a "road verge" and "river bank" landcover type.

That seems it would help transitions for river / streams which catch the eye a lot. Would allow denser/greener vegetation near water, or random water-side objects. Road verges can get objects, vegetation sound barriers, tire marks in mud etc. Verges may be useful for beaches too if the seaside was knowable somehow.

Kind regards

[wkitty42]

Maybe (?) it can be done with existing tech? A small attachment landclass area can allow a 3d waterfall particle emitter to be automatically placed by default? Texture coords & vertex attributes etc. would have to provide orientation, height, guess at flow velocity, cross-section. Don't know about mist, rainbows. Existing waterfall particles may also do mist(?). Weather tech may be able to do rainbows if a rain (droplet) volume is defined when detecting a waterfall attachment landclass.

i don't know about triggering it by landclass... AIUI, there's specific values that need to be set for each... go to PHLI and fly heading 304-305 true to the Wailing Wall as seen in the movie Jurassic Park to see an example... i'm sure wider streams can be done, too... just not automated... specifically coded... i think the same system is also used for the volcanoes

[statto]

I've got a master custom scenery, it's OSM and hand-drawn data over the vmap0 layers for Australia and New Zealand (though no custom data for New Zealand, just the vmap0 unfortunately.) Does a few things right and a lot of things poorly, but I think it's a good start. Will upload it somewhere if I haven't already.

[vnts]

i don't know about triggering it by landclass... AIUI, there's specific values that need to be set for each... go to PHLI and fly heading 304-305 true to the Wailing Wall as seen in the movie Jurassic Park to see an example...i'm sure wider streams can be done, too

That's exactly what I meant by using 'existing tech' . FG automatically placing existing waterfall emitters with parameters read from a special landclass data & shape - to auto-gen a default waterfall. That might happen in scripts that sample terrain like happens with weather & cloud placement AIUI, or from c++ engine code. I guess terrain generation could output waterfalls directly (to terrasync objects folder?) making it easier to delete and manually override for well known custom waterfalls. This way every notable drop that water flows over may get a water fall. Minor drops and steep slopes with rapids could still be shown by shaders if the flow direction vector for rivers was saved as vertex attributes in river polygons.

Kind regards

[wkitty42]

it is an idea... how feasible it is to implement, i'm not sure... i will point out, though, that without certain precautions, things like this can be instantiated and consume processing power even when they are not within range... think niagra falls flowing even when you're at KSFO and can't see it at all... this is one reason why the volcanoes are only partially done and each one has switches to turn it on/off as well as to regulate the amount of activity they have... i think it is one place that a range type sensor is used these days...

[frtps]

I've got a master custom scenery, it's OSM and hand-drawn data over the vmap0 layers for Australia and New Zealand (though no custom data for New Zealand, just the vmap0 unfortunately.) Does a few things right and a lot of things poorly, but I think it's a good start. Will upload it somewhere if I haven't already.

I think I found your work online (I still have a "Statto-Oz" directory even) but ended up regenerating from OSM and state government GDB files, most of all because I wanted to put in cliffs. Anyway, the intention is not to develop and distribute shapefiles, but to develop textures, some objects, and recipes for the shapefiles. These days OSM seems pretty good as a data source for a lot of things.

[frtps]

So I've now set up an Australian scenery repository inside an "Australasian" scenery group, in case there are those who would like to work on New Zealand and Papua New Guinea and any other islands in the area under this umbrella. The repository is currently populated with almost the sum total of my work which comes down to a couple of textures, a recipe (in Julia) for auto-generating scenery from raw OSM and GDB data and a couple of regional definition files that are in no way complete.

I will carry on gradually developing stuff at my slow pace and plan to use this repository as the place that I put my work. And it would be great if there were more contributors. Even if you do not feel moved to make substantial contributions, feel free to raise issues with the current standard FG Australian scenery, as well as any new scenery.

[Johan G]

You might want to consider adding an Australia article to the wiki. It might make it easier to find for other interested in it.