Introducing World2XPlane...

[tonywob 4,768]

Had a quick break from programming and added a few regional UK houses, 4 in all. It's amazing how far just 4 models will go, and I'm actually surprised terraced houses worked as well as they did.

 

 

 

 

Also, the zones are working well, as it has quite clearly separated out residential and industrial zones, and knowing this area very well, it also looks quite realistic. We could do with more models to diversify, the scenery.

 

 

 

[saed45 38]

tony how many models do you have and how many more do you think you need?

[tonywob 4,768]

I now have 10 for the UK, and 13 for Central Europe, all of which I've created. I'm getting a faster at doing them, and create a set of houses in around an hour or so using the textures I already have.

 

Also, on top of that, there are about another 20 houses in the R2 library which are used.

 

We need much more, especially industrial.  I take a small area in OSM, convert it, and have a look at which houses are still facades, write down the dimensions and make some houses for them. I have a good range of houses around 9-11 by 9-11 meters, but need bigger and smaller, and some L shaped, etc.. 

[alpilotx 505]

I'm trying to think up an automated way to guess if somewhere is a city. Storing a list of towns and cities seems a bit over-the-top, and also some towns might not be populated very well in OSM, and limiting such sparse areas would be a shame. 

Psst, psst .... landclass, landclass .... I repeat what I told before: add some 3rd party attribute information like landclass (or any other raster info, like population density etc.)! I do it already in my "farms and treelines" generator (which I am just in the process of rewriting  / improving for a v2 release). I use my landclass data - ok, I have the big advantage, that i have a lot of prepared stuff of that type - to decide where I do not need to generate treelines (for example in cities, in forests, in rocky mountains etc.)

[wolke85 150]

WOWser Tony here in post #121 - this looks awesome!

Simply amazing!

 

Psst, psst .... landclass, landclass .... I repeat what I told before: add some 3rd party attribute information like landclass (or any other raster info, like population density etc.)! I do it already in my "farms and treelines" generator (which I am just in the process of rewriting  / improving for a v2 release). I use my landclass data - ok, I have the big advantage, that i have a lot of prepared stuff of that type - to decide where I do not need to generate treelines (for example in cities, in forests, in rocky mountains etc.)

 

... sounds very interseting!

...

As outlined and discussed here in this thread already, but just in case saying it again: Wouldn't it be worth considering to set up/organize some kind of - sorry i do not have a proper word in english - "organized" workflow where various people interested in getting involved into and supporting theis promising looking development process can join in.

I know this might take some time to do so and it may be to early yet  - but - as there has been some very positive feedback by various people here so far and also lots of interest been shown, doing so might really be a considerable option.

But as for now - again saying:

Fantastic work so far Tony!

[tonywob 4,768]

 

 

Psst, psst .... landclass, landclass .... I repeat what I told before: add some 3rd party attribute information like landclass (or any other raster info, like population density etc.)

 

The application currently processes the landuse=* tags inside OSM, and uses that to determine what type of building to use for generic buildings. It could easily use other sources, such as ESRI shapefiles, or raster images, however the question comes down to whether it should be runnable by average users . I've also found that the corine landclass/forest information has been imported into OSM for parts of France and Spain, and these are actually causing me problems, as they haven't been broken down into smaller areas, or don't include any holes where the zone isn't valid, also other zones on top have them have to be taken into consideration.

 

With regards to using a database, I have the following issues in this regard:

 

1) If I make the program only runnable by people who aren't scared of setting up a database, etc. Then, not only will the generation run quicker (after you already have an OSM database inside postgis using osm2psql), but it would make geometric tests much simpler. Currently, the area tracking and collision tests I do are quite slow, and the only way to speed this up and reduce memory footprint is some sort of spatial database. Adding more datasources will make this situation much worse :-)

 

2) Like OSM2XP, the program can deal with large files, but this is based on how much memory available. This application stores much more than OSM2XP, as it needs to process multipolygons and areas, etc. So I don't know what the biggest file it can deal with is. I've managed to process the British Isles, which is about 650MB, although it takes a while to process. My development computer has 16GB of RAM, and an SSD, which is needed when processing large files.

 

3) For large files, I've implemented an embedded H2 database. It takes forever to populate it, but once populated, it's quite fast. Also, this gives resume support, so you can just stop your generation and continue later on without rerunning the slow information gathering routines. This is the same when using PostGIS as well.

 

On one hand, I honestly think that most simmers won't bother with the application anyway, and will much prefer ready made sceneries, e.g. by simheaven, on the other hand, once I've implemented smart exclusions and boundary exclusions, the need to generate an entire continent will disappear, and instead it will be possible and far more sensible to generate individual countries or counties without the exclusions blocking each other out.

 

 

 

Wouldn't it be worth considering to set up/organize some kind of - sorry i do not have a proper word in english - "organized" workflow where various people interested in getting involved into and supporting theis promising looking development process can join in.

 

I've set up an area to create models, and although I had two requests to access the shared folder, nobody has added anything or requested anything else since, additionally I only received a few comments on the test scenery. The problem is getting people to join in and help, and although people are keen, it isn't happening like I expected, which is why I'm trying to make models myself and keep the momentum going with screenshots, etc.

 

This is also another reason why I'm considering going over to using a database and more 3rd party sources for data, so that I'm no longer just tied down to some of the bad/missing data inside OSM. The program will still be runnable, but will require a little bit of skill to setup everything correctly, etc.

[alpilotx 505]

And remember Tony, neither is PostGIS a super / mega / fixes-all solution ... It brings its own special problems. But overall - at least this is my experience - it makes working with geodata a lot more easier (if you know some concepts of working with geodatabases), because it brings a big (and quite well tuned) set of tools with it (which then you don't need to reinvent). And the larger the dataset, the more "oomph" your computer will need ... for example when loading an entire Planet OSM, it can easily take 10-12 hours .... on a really fast and big SSD (on a HDD you would be lost entirely). But after that, you have the complete OSM extract at your hand, and can extract from it to your liking and transform it as much as you want ...

 

Should you really consider taking this way (PostGIS), let me know, as I can help you getting started (I have already fallen in many holes along this way ... no need to repeat them all).

 

And with PostGIS, you can also quite easily include other sources (like raster data etc.) ... only if you want to transform raster data (which for example I do a lot), you will need to consider another tool, namely GRASS GIS, which is very powerful when it comes to raster geodata processing.

 

Finally about the 3D models ... well, I don't want to promise too much, but I am already considering to help you a bit there. At the moment I want to finish my current "little" project, but after that, I might try to do some "objects" (as I already have a little Blender experience ... well, not really much, but enough for a few houses and texturing).

[Pascal_LSGC 298]

Hi Tony,

Thank you for taking OSM creation a step further. Your latest pictures are realy impressive!

 

 

 

As facades are replaced with objects, and things like trees are added into residential areas, performance is becoming a concern in the simulator. In Central Berlin, with HDR enabled, it was really jerky on my computer (which admittingly isn't too powerful), and I have to turn down the settings to get it decent. Although it looks fantastic outside of cities and performs well, this is now becoming a problem. The more eye-candy that is added, the worse it becomes. If anyone has any suggestions, then please let me know. This is also why it is super important to create really efficient 3D models and textures, something which I'm not too good at.

 

IMHO, facades are not yet used to their full potential. The format seems to have very effective performance thanks to instancing. We could say, the ratio "buildings variety/performance" is very high, thanks to the infinite possible footprints and wall variations. Way higher than for 3d objects.

 

The difficulty with facades lies in inclined roofs. Benny used them only with simple rectangular buildings because the height of the roof is easier to calculate. This calculation is important, because if roof height is set too high, the rendering is incorrect and uggly.

 

Why not try to calculate the maximum roof height for any arbitrary footprint ? Or try to find a simplification algorythm which would help roofs not to go "over the top" ? I know this is no easy task. Btw, I notice you already do a great job of replacing footprints with 3d object.

 

I have a few initial ideas; should you try to go this way, I could try to help develop such algorythms.

 

Furthermore, did you hear about facades v2? This is a more powerful version of facades in v10. They allow to develop a prism shape along the path, and even add some attached objects on the walls and roofs. This was intended primarily for airport buildings, and I don't know about the performance. I think this would be perfect for adding roofs to buildings with inner yards.

 

 

Pascal

[wolke85 150]

 

 

I've set up an area to create models, and although I had two requests to access the shared folder, nobody has added anything or requested anything else since, additionally I only received a few comments on the test scenery. The problem is getting people to join in and help, and although people are keen, it isn't happening like I expected, which is why I'm trying to make models myself and keep the momentum going with screenshots, etc.



This is also another reason why I'm considering going over to using a database and more 3rd party sources for data, so that I'm no longer just tied down to some of the bad/missing data inside OSM. The program will still be runnable, but will require a little bit of skill to setup everything correctly, etc.

 

Hi Tony!

 I fully understand - and apologize if my wording did cause some possible misunderstanding. I was just - maybe a bit to much - overmotivated (again) and thought that after various replies here in this thread, that - just to avoid potential double-tracked working - listing a bit who is doing what, may be useful. But sure: This only makes sense when more people join in!

So: Back to topic again!

:smile:

[saed45 38]

tony i want to help you out while i can . but unfortunately i have no idea where to start. i dont know what apps you need/ and what perimeters to use. where to get textures and ece. 

 

if you use a tool thats free and open source like camstudio and make a video of your self from beggining explaining how your process i would def put some time to help you make the app. please consider it

 

http://camstudio.org/

 

thanks

[NoComment 1]

 

 

1) If I make the program only runnable by people who aren't scared of setting up a database, etc. Then, not only will the generation run quicker (after you already have an OSM database inside postgis using osm2psql), but it would make geometric tests much simpler. Currently, the area tracking and collision tests I do are quite slow, and the only way to speed this up and reduce memory footprint is some sort of spatial database. Adding more datasources will make this situation much worse :-)

 

I personally think this is a good idea because this gives more performance in both, developing and using the tool. If anybody considers using the tool, I am sure they will spend some time in setting up the prerequisites. And if it's documented well this shouldn't be a big problem anyway. Also my understanding is, that the idea of this tool is use the best data avaiable and generate a fantastic scenery, and a GIS database will make this easier.

 

 

 

I've set up an area to create models, and although I had two requests to access the shared folder, nobody has added anything or requested anything else since, additionally I only received a few comments on the test scenery. The problem is getting people to join in and help, and although people are keen, it isn't happening like I expected, which is why I'm trying to make models myself and keep the momentum going with screenshots, etc.

 

I think to get this really started, it would be a good idea to start something like a git rep (or another central, public space). So one could enter a feature request for a let's say Italian building and somebody who sees that could submit one. It is easy to say that "everything is missing", but I think people are a lot more productive if they know exactly what they need to do. Also this would allow people, who lets say have no idea of 3D modelling, just to submit texture photos/pictures/etc (finding textures which have appropiate licences needs a lot of time but everybody can photograph their own house ).

Also clarifying what exactly is need for submitting a model would be cool. I have worked with blender for years (mostly VFX stuff, but I can do a low poly house ) and could easily create buildings or even a tutorial. BUT I don't really know what kind of a file you need (and I am writing exams for about 5 more weeks, so I am busy till then).

[tonywob 4,768]

 

 

I have a few initial ideas; should you try to go this way, I could try to help develop such algorythms.

 

Thank you for the information Pascal. I have looked into extending the facades, and also using the v2 format, but I found the documentation very hard to follow, and without any sort of visual editor and examples, I gave up with frustration. I decided to go down the object root. However, I haven't ruled out facades, as they work really well in cities, especially for multipolygons, office blocks and large industrial buildings. If you understand the format and can offer help or facades to use, then please let me know :-). Also I have trouble getting the sloped roofs to work, even on very simple square buildings, so if you know why/how to get them to work, then again please let me know.

 

 

 

I think to get this really started, it would be a good idea to start something like a git rep (or another central, public space)

 

I thought of using github to host the objects, textures, etc. But I thought people wouldn't want the hassle of installing a git client, etc.. I think however, this is the way to go, and I'll setup a repository, and check in the current models I have. Github also includings a basic ticketing system, so this is perfect for your suggestion, and I or others can open tickets with models or photographs we need. 

 

 

 

tony i want to help you out while i can . but unfortunately i have no idea where to start. i dont know what apps you need/ and what perimeters to use. where to get textures and ece. 

 

Agreed, perhaps I can knock up a simple PDF in the next few days with some screenshots of the process I am using to create a house using sketchup. Also, once the repository is up, there will also be plenty of examples to look at. I'm not a big fan of doing videos, and with a PDF, you can easily see pictures and instructions for the various stages.

[Humpty 751]

@Tony I am just waiting for the brilliant stuff you are doing to be final. honestly dude awesome.  

 

If you can manage the trees region wise WOWOWWOWOWOWOWOW. 

[NoComment 1]

 

 

I thought of using github to host the objects, textures, etc. But I thought people wouldn't want the hassle of installing a git client, etc.. I think however, this is the way to go, and I'll setup a repository, and check in the current models I have. Github also includings a basic ticketing system, so this is perfect for your suggestion, and I or others can open tickets with models or photographs we need.

 

I agree git can be quite a hassle. But alone the ticketing system will show others which works needs to be done, so this would be a great benefit. Also all suggestions which are far in the future (there were quite a few here and at x-pilot) can be entered as a feature request, so we don't forget them. Also if somebody doesn't want to install a git client, I guess they could email it to somebody who is able to put the file into the rep.

 

 

 

Agreed, perhaps I can knock up a simple PDF in the next few days with some screenshots of the process I am using to create a house using sketchup. Also, once the repository is up, there will also be plenty of examples to look at. I'm not a big fan of doing videos, and with a PDF, you can easily see pictures and instructions for the various stages.

I think this would enable many to contribute since 3D programs generally have a rather steep learning curve. I will contribute a blender tutorial as soon as I have some free time ...

[alpilotx 505]

By the way: there are already some fantastic free texture sources you might check out to "puzzle together" some stuff for your buildings (I have already used a bit from there):

http://www.cgtextures.com/