Jump to content

Oscar Pilote

Frozen-Inactivity
 View Profile  See their activity

  • Content Count

    37

  • Donations

    $0.00 

  • Joined

  • Last visited

Posts posted by Oscar Pilote

  2. Posted

    The precise reason why X-Plane's ground handling is fundamentally wrong (here and here). The simple root cause has never been addressed as far as I know, only piles of hacks to try to fix the unfixable. The details were sent (with the softest gloves) to Austin last autumn, but he never answered. If people like Jan believe it is mostly ok as is, then for sure it is not gonna change any time soon. But from the physics point of view it will remain just an ugly hack around a neat solution of classical mechanics known for more than two hundred years (and readily available in third party open source code...).

    • Like 2

  5. Posted

    On 05/11/2017 at 1:36 PM, alpilotx said:

    So, to get the best results, someone would need to capture "side" images of the steep slopes and apply them in the right place .... Yes, doable, but definitely nothing for the faint hearted (and neither something which can be easily automated). But a scenery like this would be the ultimate experience fur sure ...

    This is called oblique imagery and is e.g. what Google Earth 3D is using. Once available it is not more complicated to introduce them in the scenery with repect to orthos, only the azimuth (usually 0,90,180 and 270) and zenith (usually 45) angles need to be known. Google used to propose them through their gmap API (although out of big cities most were fake ones - satellite orthos distorted), now it seems they have abandoned it and only serve their already textured (and highly protected!) GE 3D mesh, where available.  Of course oblique imagery does not replaces a good 3D mesh (the video was a joke, buildings are not 3D in the mesh, unlike GE), but for texturing steep cliffs they would be quite handy.   

    • Upvote 1

  6. Posted

    On 05/11/2017 at 8:22 AM, Susu986 said:

    Yes, you can use HD Mesh to build overlays for Ortho4XP tiles, and use those overlays (which is basically the autogen buildings, forests, and roads systems on top) with Ortho4XP. However, you won't benefit from HD-Mesh' high resolution terrain mesh at the same time while having orthophotos

    True. Yet, the mesh being different doesn't necessarily imply that it is not as good. Here is another side-by-side comparison I wrote back two years ago or so (pay attention to the shapes, not the textures).  Note however that the HD v3 and UHD v1 triangle counts were penalized by a nasty encoding bug (related to the 32 -> 64 bit change, most of the nodes were duplicated) which was also the cause of Laminar's initial "Global scenery release size problem" (if you read about it) for XP11. That bug has been removed now and HD v4 should not suffer from it.  

    • Upvote 3

  8. Posted

    Dear Brian,

    Sorry for the misunderstanding, I truly thought you were not aware of that project, I am reading the forums (only very occasionally for avsim) for but not to the point to know every one and I don't remember having read from you already. Perhaps for the same reason I don't understand the meaning of your sentence about the posters and Forkboy2 disappearing (?). It probably doesn't matter. What I meant was that if people have the choice between a free and a paid version of "roughly" the same product (if you make that incorrect that's perfect! but I was told Forkboy2's work is high quality too) chances are they will go for the free version.

    The tree project encouragement was not fake, to my opinion this is one of the few points where X-Plane is really lagging behind (although I have nothing to compare it with).

    Again my apologies if you had the impression I wanted to discourage you, it was not my intention at all.  If your signature is correct and you come to Fontainebleau, I'll be glad to share a bouldering session, the best season just started (and mushrooms too :-) 

    Cheers

     


     

  9. Posted

    Dear Brian,

    It is a bit unpleasant to tell but you seem not to be aware of this other project which is also based on NAIP imagery and 10m elevation data.

    https://forums.x-plane.org/index.php?/files/file/36327-us-orthophotos/

    That one is not concerned with the overlay part though, so if the commercial nature is important to you it is probably a safer bet to focus on trees (which is a very nice project per se)

  10. Posted

    15 hours ago, Richdem said:

    There is supposed to be some serious optimizing in the pipeline.  One thing I have found is that disabling the extended DSF in the settings file does make the VRAM usage with Ortho4XP scenery way more manageable.

    I think I have singled out the guilty guys. If you look into the textures directory of any tile you will see the .dds orthophotos but also the .png masks (if sea is present on the tile or if you use use_masks_for_inland). It turns out the BORDER_TEX textures (i.e. the masks) are never mipmapped by X-Plane, and therefore they cost 16Mb of VRAM each (4k * 4k * one 8bits channel) even when they are 150nm away. If you are NOT using use_masks_for_inland (which I recommend) their number is limited enough in general and even with extended_dsf the additional VRAM need is of the order of 1Gb at most. With use_masks_for_inland you''ll have something like 15-20 of them for each of the 12 tiles, and therefore a burden of an additional 3-4 Gb in the worst case.

    In the "stock" X-Plane BORDER_TEX is used with small generic png files like texture gradients or similar; this is the reason why they did not implement the mipmapping for them. Ortho4XP is using individual 4K textures so the issue is very different. I have e-mailed Ben a few days ago, he and Sidney will try to make it for 11.10 (BORDER_TEX will have a LOAD_CENTER directive as BASE_TEX already has). This alone should cut the VRAM need of masks to one or two hundreds of Mb (be it extended dsf or not).  [on the other hand they will get much heavier on disk size (22Mb dds dxt5 each vs xxx Kb png), but still "negligible" compared to the orthophotos].  

    Also (this is already the case), because orthophotos are mipmapped you shouldn't see a big impact of texture zoomlevel on VRAM usage (only the closest ones need full resolution), you may check it in the sim but I would say that the NET impact of dds orthophotos on VRAM is
    of the order of 2Gb at most (at highest quality with compression - by NET I mean excluding masks, e.g. with water_option=1,  and excluding X-Plane's base need, almost 1Gb in my case as seen in the middle of nowhere).

    Hopefully my GTX 970 4Gb will be good for service an additional couple of years.

    • Upvote 2

  12. Posted

    (a copy/paste from a post on x-plane.org forum)
    Regards
    Oscar

    ----------------------

    I have uploaded under the following link mesh files for 30 tiles in England (from South up and including +53).  

    https://drive.google.com/drive/folders/0B3vPD6IeCB9zbU5JWmlWNlpfeDg?usp=sharing

    Elevation data are (a jigsaw) from Ordnance Survey (2m) and Environment Agency (50m), two Open Government Licence datasets, in particular almost all of the coast is available under 2m (which I anyhow downgraded to 5m for convenience and easier gluing).
    Vector data (OSM) is used a posteriori to tweak the raster elevation data in a tentative to improve road banking and to avoid airport flattening without letting them too bumpy. 

    I have not tested the mesh files except for Scilly, Cornwall and Plymouth (which are great places to fly, I found !). I'll be happy to get some feedback on airport smoothing, this shot is one same parameter for all and there is surely defects. In particular if you find one that turned non flyable please let me know so that I can try to figure out what could be the cause and how to overcome it without making a ad hoc (I should say haddock :-) rule.  

    -----

    To use them :

    Unzip a(ny) mesh file in its corresponding zOrtho*** directory (replace the older .mesh if existing). In Ortho4XP, select the corresponding tile (and custom build dir if not default), provider and zl (or press Load cfg if just adapting a tile you have) and then DON'T touch step 1 or 2, just Step 3 (and 2.5 if you wish to redo masks).

    -----

    Tony's GB Pro 2 is certainly a very good choice of overlays for using on top of these, at least this is the one I chose for testing and it looked superb there.

    Cheers from the continent

    • Upvote 3

  13. Posted

    Just to mention (even if they didn't have to do so) that they contacted me last December

    to ask and tell me about their intention. I forgot about it but checked after a couple of people

    wrote me these days.

     

    @Pankaj I hope you'll agree that the "lazy" customers as you call them are only one half in the equation,

    people involved in the free (software) community also tend to believe they take their part (the debate

    whether open source would kill part of the economy as long been settled by the facts).

  14. Posted

    Hi,

    Here is what I would find good to have :

     

    1) A public repo for .mesh files. The .mesh file already contains all the relevant vector and raster information, and Step 3 can be performed straight from it by adding the choice of ortho

    covering. And there is no legal issue with the information it has if it is derived from public data. They would be accompagnied with information on the vector and elevation data that were used to build them.

    New free sources of high res DEM and vector data are poping almost each day, and local users know better where to look at them. But when a high res DEM for 1 tile can be more than 1Gb (and requires quite some time to process/reproject etc), the resulting .mesh once 7ziped is only a few tens of Mb and is ready for use. Perhaps it's wise to think of a few mainteners that would receive the uploads and accept or refuse them based on quality and/or already existing stuff.

     

    2) A public repo for free imagery already cut in the good format and possibly "uniformized" at least to the extent of the upload.

     

    If there is sufficient added value in 2) then people will stop downloading from the big two (although I am not too much worried about these services cutting us access because my intuition is

    that local sources will end-up replacing them almost completely)

     

    3) Additionnal server(s) coming in support of zonephotos; Pascal who is running it can made it transparent so that there would be only one website entry but more than one server below. Actually

    he also told me long ago that something like 1) he could implement it on zonephoto as well, he's presently writing an app to replace the in browser java applet.  These servers could go on

    playing border line as they do  now, Pascal did indeed contacted Bing and Google long ago and they were only a bit vague in their answer.  The present cost of the zonephoto server is something

    like 40$ a month (I think its BlueHost).

     

    Now that said, I would be terrible at trying to manage even parts of such things, so please go on for the best !

    • Upvote 2

  17. Posted

    In case of added value there is nothing wrong with trying to make a profit, we all need to make a living from something.

    But I replied mostly because the argument about "collective efficiency" can be rather counter-intuitive.

    The point regarding copyright related to Digital Globe or other license holders remains certainly true, I feel much more comfortable when I find a good open-data imagery.

     

    I think apes don't use +1 or -1 either, that was invented by those barbarians of the Roman empire in the circus :-)

    Season greetings to all of you

    • Upvote 1

  18. Posted

     

    Personally, I also dislike that whole concept of millions of apes sitting behind their machines and crunching the same tiles to XP scenes over and over, with the only difference being perhaps the ZL. It's a waste of manpower as well as a waste of bandwith.

     

    If one wants to compare a centralized versus decentralized tile creation/distribution the facts are :

     

    1) (everything else being equal) it does not take longer to build one tile than to download it from a centralized source, it is even usually quicker because providers have larger bandwidths.

    2) decentralized processes are as automatic as a file download, resuming to a few user clics, so manpower usage is comparably close to zero.

    3) total bandwidth is actually smaller in a decentralized process, because jpegs are roughly 2.5 smaller in file size with respect to dds (and still roughly 1.5 smaller when dds are 7z compressed). 

     

    The main difference is the channel supporting the bandwidth of course, and that has to be taken into account when the download is not welcome.

    Regarding EULA : the best imageries I know of are governmental and an increasingly larger number of them are copyright friendly for non commercial (sometimes even commercial) use. In Europe countries are enforced to do so according to the INSPIRE directive. In the US the USGS imagery has long been copyright friendly.

     

    That said : STOP doing ZL19, in 99.9% of the cases it is purely useless, a true waste of everything.

    Most of the time :

    - optical resolution is close to ZL18 at most, sometimes even ZL17,anything higher is just oversampling.

    - above a few hundred feets lower mipmaps will take the place anyway, so your top resolution will

    not even be loaded.

    For VFR use ZL17 with notches of ZL18 in the places you know you will fly very low. For anything

    above 2000ft ZL16 is already almost optimal. For airliner altitude anything above ZL14/ZL15 is simply not loaded as a mipmap.

     

    Mankind still have a few things to learn from apes, perhaps even courtesy.

    Regards

    • Upvote 1

  20. Posted

    I feel it is not working well with quite a few hgt files, don't know if the problem is in the hgt files.

    What do you mean by this.

    I you mean the message with "no data zone" that I make printing on screen this is, at is says, because

    the hgt file has no data zones (encoded as altitude=-32768). There are many tile with this on the

    viewfinderpanorama website, but the zones are tiny so the averaging process that I do for those

    should be fine in most cases.

    If you prefer the .tiff files from gdex do not have no data zones (at least I have not seen some with)

    but viewfinderpanorama even with a lower resolution is the prefered  choice of a number of people.

    Does this tool has this problem as G2XPL does?

     

    http://forums.x-plane.org/index.php?showtopic=82010&p=910530

    I do not have those airports but 99% percent sure the issue wouldn't be on g2xpl or ortho4xp side.

  21. Posted

    In principle I could make the conversion process quicker by multithreading it (say after the download process is finished), it you look

    at the resources manager you will see that only one core is used when only the conversion remain.

    An alternative is to tick the "skip convert" option and to batch do it later (multithreaded or not).

    Also, for certain WMS servers the download process is not faster the the conversion one, sometimes even slower.

  22. Posted

    I can't seem to see your png but open the file Ortho4XP.cfg and

    you'll see a place where you can adjust saturation, contrast and brightness

    on a per provider basis (think of the numbers as percentage which you add (positive)

    or substract (negative) to the originals).

    Some users find textures washed out when they are not very constrasted and

    saturated, others like you (and I) prefer them more natural looking; with these

    parameters anybody can adapt to their preferences.

  23. Posted

    In ortho4xp.py in line 3194 a blank before -b was missing.

    Thanks for that, I did a bit a cleaning by cutting long lines before uploading

    and that way I broke it...

    I am so stupid that I also uploaded it with the error on the .org...

    I just don't get the point of flying over flat world it's to 90s for me.

    One point could be that the 3D buildings which you will then put on top of it

    sometimes look better with such a flat world beneath, like e.g. here.

    Orthophoto vs landclass is a matter of taste, I personnally like both.

  24. Posted

    Hi Hans,

     

    In principle nothing changed at the level of Gimp or the masks between this version

    and the previous one. I will reboot in windows later today and try to reproduce it.

    I tested it a few weeks ago on windows 7 and it was working, and indeed yes I had

    to go through a batch file (only under windows, I otherwise had issues with "escaping" too

    much quotes).

     

    Regarding you previous message, I must confess I have no idea of what an ESP platform is (?),

    but I am sure you'll tell me !

×
×
  • Create New...