Posts by crispy136

    Had a look in the download folder and surprisingly both folders had the same number of files, but slightly different disk size used. I tracked it down to one jpg. AeroScenery seems to have downloaded something, but when I try to look at it, it is not a valid jpg. I assume it created the name but then had nothing to put in it


    FSET used a queue system for any images it wasn't able to get straight away, although it used to end up with quite a few in the queue not just one. So I don't think it did any retries on the first run. Once it finishes all the other images, it returned to the images in the queue and did a second attempt. I never had an instance where it failed to get all images, so it seems to just be a timing issue.

    Hi Nick, I noticed I am getting a few masked tiles in 0.6 I wasn't expecting. Tracked it down to missing source imagery in the stitched image. Rerunning the stitch process didn't make any difference, so I am guessing that the source image wasn't downloaded. When I reran the AeroScenery process again in full on the same map tile, same zoom level, all source images were downloaded and no mask ttc's produced. Have had it happen on Bing and Google and on level 9, 13 and 14 map size selections. While it is quick an easy to rerun a level 14 or 15 map tile, not so quick on a level 9 sized map tile.



    I've been having a poor experience with GeoConvert with AeroScenery

    These are the following AeroScenery combinations I have found work reliably on my PC and produce good quality results when viewed in FS2. Times are only indicative of running time based on my PC's specs.


    1. For map areas with no airport - gives VFR quality above 6,000ft

    - Select map tile 9 & zoom level 15 & output levels 9, 11 & 12 (approx. 13mins & 1.1gb disk space per map tile selected)


    2. For map areas with an airport - gives VFR quality above 3,000ft

    - Select map tile 9 & zoom level 16 & output levels 9, 11, 12 & 13 (approx. 60mins & 4.4gb disk space per map tile selected)


    3. For town areas near an airport plus anywhere under the aircraft flightpath - gives VFR quality below 3,000ft

    - select map tile 13 & zoom level 18 & output level 14 (approx. 9mins & 790mb disk space per map tile selected)


    4. For airport inner boundary (optional) - sharpens textures at ground level

    - Select map tile size 14 & zoom level 19 & output level 15 (approx. 3mins & 151mb HD per map tile selected)


    For a level 9 size map tile with an airport, I run 3 AeroScenery sessions (2, 3 and 4 above). I never run more than one instance of AeroScenery at a time, but this will produce as many GeoConvert sessions as map tile selected. Both CPU and RAM get maxed out with multiple GeoConvert sessions, but leaving the PC alone until its finished works just fine. I usually run AeroScenery with download, stitch, TMC & AID and GeoConvert all selected.


    I also found that there is a limit to the number of GeoConvert sessions you can run at one time. For my system I can run 8 of 1 above and 10 of 3 above. No 2 above is quite labour intensive for GeoConvert, so I usually run these with just 1 or 2 map tiles selected at a time. As 4 above is just for the airport inner boundary I never need to run more than 4 map tiles together anyway. You could test just one or each above to see how your PC specs perform.


    You haven't indicated what AeroScenery values you had selected. If going for the quality everywhere approach, with a selection of map tile 9, Zoom 18, and output levels 9, 11, 12, 13 & 14 selected, you may be asking your PC for too much. My testing indicates that this combination of values in AeroScenery would result in a processing time of > 12 hours and HD space required of > 100GB on my PC. So I stay away from this combination and only target quality where it is needed.


    There are 3 scenarios I encountered that failed to produce any output:-

    1. using zoom 18 or below with output level 15.

    2. running with masks off with a stitched image with a transparency. Unfortunately AeroScenery doesn't indicate if it created a stitched image with a transparency.

    3. Having too many GeoConvert sessions running at once. I was running 9 of 1 above and it didn't produce any data for the 9th map tile, but no errors reported.

    Interesting result and needs some debugging by any willing volunteers!

    Have attached the results of the map 13 sized tile extracted @ zoom 18 (0.597 m/pixel) in AeroScenery and download res 0 (listed as 0.5 m/pixel) in FSET. Have attached everything except the image files. AeroScenery image file was 8192w x 8704h png, while FSET image file was 6146w x 7258h bmp. TMC files used by both processes had the same values.


    In the FSET zip you'll find a couple of extra files.

    1. txt file (not used, but produced by FSET)

    2. inf file (produced by FSET but only used only to create the tfw file)

    3. tfw file (created from the inf file)


    Looking at the scenery tutorial the tfw file has the pixel steps in it. FSET doesn't produce the tfw file and that has to be done manually or by using a little exe someone developed, before GeoConvert is run. The tutorial gives no details of the purpose of the tfw file, only that it is part of the process.


    I also noticed that AeroScenery generated the aid file, but in the FSET process, the aid file is generated by GeoConvert. So it would appear the aid file is an byproduct of the GeoConvert process, not an input. Perhaps this is the problem, and AeroScenery needs to produce a tfw file not an aid file. Only guessing.


    AeroScenery top, FSET bottom. Note misalignment with Fscloudport runway.


    FSET - both top and bottom

    Files

    • Aeroscenery.zip

      (1.32 kB, downloaded 11 times, last: )
    • FSET.zip

      (2.98 kB, downloaded 10 times, last: )

    Found this while doing some source image quality comparisons. I think this may have been mentioned before, but couldn't find the post. Looks like the higher the zoom the more accurate the placement of the tiles. Looking at the airport runway, zoom 19 gave the best match but was still a little out. I repeated the test using map size 13 tiles and got a similar result using either level 14 output or level 13 output. In case there was some issue with GeoConvert I also did an extract using FSET, but encountered no problems. I used Google for the AeroScenery extract but the tiles I had from FSET were Bing, but I don't think this was an issue.


    AeroScenery - map size 14 with GeoConvert output size 14



    AeroScenery - map size 13 with GeoConvert output size 13



    FSET - map size 14 with GeoConvert output size 14



    I found that 2 or more simultaneous GeoConvert sessions running maxed the CPU, rendering the PC unresponsive until it had finished. However left alone it finished without causing any problems for Win10. Also that when I was running 9 tiles, GeoConvert didn't produce any output for the 9th tile, although when I reran the GeoConvert step for just that tile it worked. I didn't have any crashes with 20gb RAM, but never exceeded 9 at once. I also noted that I could run more level 14 tiles at once than level 9 tiles at once. I don't think GeoConvert was ever designed to have many instances running at once.

    We have received legal opinion on the sharing of ortho imagery and, with the exception of using USGS data, it is illegal to use Bing or any other source to share such files on a public medium for download.

    So was that opinion sought from Bing themselves or from a 3rd party like a lawyer who then looked at Bing and Google's EULA's. The response I got was by asking Bing themselves, I just contacted Bing's email address for map licensing, so their response is most likely from the US. However if there is any doubt, then the best advice is always not to distribute anything.


    In my case Its probably all a bit academic now, as I have found that reasonable quality scenery would result in data that is too large to share. But with new tools like AeroScenery, users will soon be able to easily create their own scenery. To that end I have done a bit of testing an come up with 3 different combinations of settings using AeroScenery, that would give different "effort/size vs quality" scenarios that would allow users to pick the best option that suits their FS2 needs and computer resources available.

    Perhaps their response was positive because I mentioned in my request that the use was "not for profit"and that the distribution was to others with the same hobby . I'm sure you could argue that the email from Bing giving permission to distribute imagery created from their maps constitutes a written license. After all, it came from the Business Development Manager of the Bings Maps Licensing section, so the response was from a person of responsibility in the appropriate area.

    Hi Nick, This sort of problem is definitely of GeoConvert feature. Try a different tile at the same settings and it works. Try a different zoom level and the same tile and it works. AeroScenery content is all present, but GeoConvert fails to produce any data. I have even had instances where I am running the extract for levels 14 and 15 but only level 14 is created. Same thing masks on or off. The GeoConvert screen says its doing level 15 tiles but produces none, and no errors appear in it log. It could be something to do with how close the source image is to the FS2 tile size. One tutorial for the GeoConvert helper tool mentioned that its "snap to lvl 9" option can fix these problems of missing output, although the snap to feature relates to level 9 tiles but when creating levels 9 and up together.

    I’m thinking we will be having a totally different conversation or at least totally different results once we have USGS hd scenery available for testing. IPACS may reduce the quality of the scenery when looking at it in FS2 but, the quality of the base images are far superior to most of the Bing and Google free images in the U.S.

    Unfortunately better source quality doesn't translate into better images in FS2. Also doesn't USGS map data just cover the USA? Their terrain data is USA only. There's a whole rest of the planet out there, even if some of us are standing upside down. ^^

    Level 9 is ok seen from a cruising liner at FL300+, but not for VFR.

    If you create level 9 and 11 together using best image quality for level 11 area, you would see level 11 imagery below you. Level 11 tiles become visible at 40nm, so level 9 detail is only visible in the distance. At FL300 you would still be seeing level 11 tiles within 40nm. I have found that even level 15 tiles, if you have enough of them, are visible at FL600, as tile draw distance is not affected by altitude. You just have to be looking straight down. I did a quick test with level 9 & 11 scenery by itself and your right, it is not at its best under 3,000ft. It gets better as you climb and at 5,000ft is acceptably realistic. So if you wanted to do VFR flying below 5,000ft I guess you would need to create the area with a minimum of level 12/13 and depending how low you want to go, maybe even level 14. I not a pilot just a simmer, so have no real world experience of VFR.

    If you intend to cover a place with variable resolution, it should be all based on the same source resolution, unless you are pretty confident that the source images show the same thing whatever the resolution.

    That's a good point. But even using one image quality source you still get some quite noticeable imagery contrasts from place to place. This is just the limitation of using available imagery.


    It seems that when using tool like AeroScenery, or FSET for that matter, you want to minimise your effort while maximising your quality of output. What I want form FS2 scenery is less detailed imagery everywhere for flying above 5,000ft, then more detailed as you near an airport area and the finally the most detailed imagery when you are on final approach. I thought those 3 requirements neatly dovetailed with GeoConvert level groups of levels 9 &11, levels 12&13 and levels 14 &15. You would also use an image source that maximises quality for the higher level of the pair and then let GeoConvert dumb it down for the lower level of the pair.


    I think we will always have to live with the possibility of levels 9 & 11 will sometimes have different textures to others, as using imagery quality higher than level 15 simply generates huge amounts of data for little benefit. However if we can find a quality sweet spot that would let us generate levels 12 to 15 using the same source data that would be a bonus, but I'm not confident this is possible. The beauty of AeroScenery is that anyone can generate all levels using one image quality source if the have the time and computer resources to do so.


    I understood AeroScenery is a tool for everyone, not just for the experts. In order to make it usable for those with less computer skills (and computer resources) we have to make it easy enough to be worthwhile. To do this we just need a set of rules we can apply to an "easy mode" in AeroScenery. This is what I was trying to achieve. However if that is not the purpose I still want to know the sweet spot for my own scenery design.


    So I still think FS2 final image quality is what should drive the image source quality levels used, but maybe we can do it for more levels at a time to avoid the problem you mentioned. I'll keep working on finding the sweet spot. For level 9 & 11, I already have found image quality level 15 is it, as you lose quality using level 14 but gain nothing using level 16 on level 11 FS2 grid. I happened to pick New Orleans for my testing and found no source image variations between 14, 15 or 16, but this may not be the case everywhere. I think we'll go crazy if we think about this too long.

    Lastly, there is one additional twist here. This twist applied to FSET and probably applies to Nick's tool as well: with certain imagery sources, when you get to the highest 'zoom levels', they don't actually increase in true resolved spatial resolution.

    Yep I noticed this issue myself using Bing data. I first noticed in FSET as it cycles through the images as it downloads them. When I redid an area using a supposedly better quality I noticed that the images cycling through the screen were more grainy. It is likely also that some areas have higher quality images than others, even though both areas have the same download resolutions available. I found that in the area of Australia I was extracting, there was no quality difference between FSET levels 0 and -1, just bigger downloads.

    I am currently doing some comparisons of what gives the best image quality in FS2. Looks like image download level 20 need never be used. And it could be we are using download images way too detailed for the various levels in FS2. Below are few shots of a GeoConvert level 9 run at both 15 and 16 download image quality. Once again source images are different but the output is not. Hopefully after a few extracts and comparisons I'll be able to work out what the max download image quality is required for a particular GeoConvert level. Once we know that then we might be able to create larger areas in less time.


    Its the stats for the level 9/11 grids that are most interesting:-

    - level 15 quality downloaded 4225 images and required 782mb of disk space (804mb free required)

    - level 16 quality downloaded 16641 images and required 2606mb of disk space (2697mb free required) and all this for no improvement in FS2.


    In both cases both the output ttf files were about 35mb. Don't forget a level 9 tiles is only 30nm x 30nm, relatively small in the FS2 world. So extrapolate that out for the HD space required for a bigger area and you'll see that using the optimal download image quality becomes important. While storage is cheap, all that has to move through your PC memory as its created.


    FS2 level 11 tiles - 15 on left 16 on right



    Stitched source files - 15 on left 16 on right


    I think the image you showed me was from the stitched image file. That doesn't equate to better quality in FS2 as GeoConvert compresses the image. My images were of the converted files ready for FS2.


    Below are the 2 source images that generated the FS2 tiles. Once again 19 on the left and 20 on the right. In my post above the FS2 tiles showed no difference between source image quality or 19 and 20. However in the source image data below used to generate those tiles, there is a noticeable difference between 19 and 20. Its really what appears in FS2 that is important.