AeroScenery Beta - Help With Testing Required

  • Hello Nick,

    thank you very much for version 0.5 of AeroScenery. It already is a very good product, which the community eagerly awaited. I like it very much!

    What already works good for me, is converting big areas in Level 9-14 (1m) without hang up. Also showing different Map types works fine.

    Currently I have a problem, that the "Show Airports" does not show a single one.

    And installing the converted files into the configurated AFS2 folder did not hppen.

    Also "Open in Map" does open google or bing, but it does not position to the correct location (also did not in earlier versions).

    Some comments to "Level 14 - Smallest":

    I converted San Francisco airport. 4 separate areas were needed, producing only one single (stitched) PNG file each. It is quite difficult to edit these because they are quite burried inton the folders. Can you establish an exeption that you do not split these over 4 folders so that all ongs can be found at one location?

    Cheers, Thomas


  • Currently I have a problem, that the "Show Airports" does not show a single one.

    And installing the converted files into the configurated AFS2 folder did not hppen.

    Also "Open in Map" does open google or bing, but it does not position to the correct location (also did not in earlier versions).

    Hi Tom,

    Thanks!

    The airports only show at higher map zoom levels (it adds too many markers otherwise). "Show Airports", then zoom in somewhere populated and let me know if you see them. Failing that check the log (which is now in My Documents/AeroScenery) to see if there were any errors when populating airports.

    I can't do the install feature until IPACS improve GeoConvert so I know when it's finished. I forgot (again) to grey out the option. Ooops.

    Strange that "Open In Map" doesn't work properly. Can you give me a google maps link and AFS tile name of an an area that doesn't work properly and I'll try to debug.

    I converted San Francisco airport. 4 separate areas were needed, producing only one single (stitched) PNG file each. It is quite difficult to edit these because they are quite burried inton the folders. Can you establish an exeption that you do not split these over 4 folders so that all ongs can be found at one location?

    Probably not on that one unfortunately. It's all been written to put things in individual folders per AFS tile. The level 13 and 14 selection side was a bit of an after thought once I realised how long it would take to download a level 9 square at zoom level 20.

    Those 4 separate areas might correspond to one level 13 tile though, so that's one option. I could add a level 12 selection size easily too.

    AeroScenery - Easily create photoreal scenery for Aerofly

  • I am leaving on vacation in two days so I am going to wait until I return in 2 weeks to download and try the new v 0.5. I will probably miss the South Florida DLC and the TE Netherlands debuts but, I will follow the forums with my iPad.

    For excitement I will try to fly a helo through the volcano eruption - I hear it is too dangerous to view from a cruise ship. =O I will also try to snap a few pictures of PHNL for our custom scenery builder. ;)

    Regards,

    Ray

  • Nick, a minor V0.5 bug, when downloading several squares, the "Current Action Progress" bar doesn't reset after the first square is stitched - while the 2nd square is downloading the progress bar stays at 100% but resets when the 2nd stitching process starts.

  • Hi Nick

    I did some testing and comparison with the various files created by the manual FSET and automatic AeroScenery processes and after a few false starts I think I have an answer.

    I used the same FS2 level 13 grid in FSET and Aeroscenery and set about to produce the output, then compared the results. Both TMCs were run through Geoconvert with Mask enabled.

    Output files created- FSET = 1, AeroScenery 3 =3. Note all images below show FSET on the left and AeroScenery on the right.

    The extra ttc (70a0) created by AerScenery was a strip across the bottom of the tile, hence the additional mask file needed. Looking at the source stitch file, the texture is from the top of the stitch image, so it in effect fits on the top of full tile created. Look at the enlarged image below to see the strip as the thumbnail process chops it off. Running the extract with masks off would have seen the mask tiles not created. I had found in my earlier scenery creation attempts that when a source image spanned multiple grids in FS2, and you are running GeoConvert for that grid level with masks off, then Geoconvert discards those tiles that would have resulted in a partial tile (mask tile). Same thing occurs with transparencies.

    TMC comparision - no difference here

    AID files - difference in lat of the top left coordinate, the first clue I think.

    Lastly I compared the source graphic files created from the image data that is read by GeoConvert. What I found was the AeroScenery stitch file was taller than the compiled BMP created by FSET. The AeroScenery stitch image has a horizontal strip of image at both the top and bottom compared to the FSET file. In the AeroScenery stitch image the extra strip at the bottom is roughly double the size of the extra strip at the top. In the image comparison below note the small creek at the bottom right of the AeroScenery image, and the small oval at the top. Both these are missing from the FSET extracted image. The width is the same in both.

    Source image comparison - note extra height in the right image created by AeroScenery

    When running GeoConvert (using an identical TMC file for both remember) the extra strip at the bottom of the AeroScenery image appears to be ignored by Geoconvert. But the extra strip at the top of the AeroScenery stitch image appears in the additional files created by the mask option. I am assuming that as the FSET source image file more closely matches the FS2 tile's width to height ratio, no mask files are created as there is no additional image area to be cut off.

    Raw file comparison of the same output file produced by both processes - both are identical except for some colour variation. The top and bottom of the AeroScenery image is discarded by GeoConvert resulting in the same output image as from FSET.

    So there appear to be 3 courses of action.

    1. Do nothing in AeroScenery and just run with masks off. This will discard the extra portion of the taller AeroScenery image. The top section of the image used in the mask file has a distinctive oval in it, so I was able to confirm that this strip also appeared as the bottom section of the full image of the tile above when that grid was extracted. So using masks off and discarding the additional image data will not compromise the scenery output accuracy. But using masks off will cause problems where transparencies are used.

    2. Do nothing in AeroScenery and run with masks on. This would mean that FS2 has 2 image tiles created for the same data for the same level. So the images in the strip across the mask tile will also appear as part of another full tile. Not sure if that has any impact, but I did read in another thread that mask file sometimes cause visual issues in FS2. Can't remember the issue though.

    3. Modify AeroScenery to reduce the overscan in the stitch file. Masks can be left on to allow for transparencies and this will not impact the output file creation as it does now.

    The image below shows the size and coordinates used by FSET that produced the FSET equivalent of the stitch file. Perhaps you could compare these details with that used by AeroScenery.

    Just as a side point, but I had to run the FSET process a couple of times to try and get a good match of the respective output file resolution. I had to try different download sizes in FSET to match the zoom level in AeroScenery, since they describe the image quality differently. I ran 3 extracts in total at different image quality levels and in just one of those, the FSET data produced mask files. However when I looked at the mask file it was just a single line of pixels across the bottom. So FSET occasionally has some very small variations that produce mask files, but they are rare.

    I have been looking at this so long my eyes are starting to bleed. Time for a break.:)

    Thanks, Chris

    Win 10 64-bit, 24GB RAM, i5-9400F @ 3.9, 6GB Nvidia RTX-2060

  • Hi Crispy136,

    Nice test and report, thanks !

    AID files - difference in lat of the top left coordinate, the first clue I think.

    Note that you have different pixel sizes in X and Y between FSET and Aeroscenery sources. The coordinates will vary depending on the settings in each tool, but the pixel size should be the same if both use the same raw image resolution.

    Raw file comparison of the same output file produced by both processes - both are identical except for some colour variation. The top and bottom of the AeroScenery image is discarded by GeoConvert resulting in the same output image as from FSET.

    There's actually a difference: look at the right edge of your pictures. Top, left and bottom edges are similar, but the right edge isn't on the same place in both pictures...

    There's probably something to investigate here...

    Cheers

    Antoine

    Config : i7 6900K - 20MB currently set at 3.20GHz, Cooling Noctua NH-U14S, Motherboard ASUS Rampage V Extreme U3.1, RAM HyperX Savage Black Edition 16GB DDR4 3000 MHz, Graphic Card Gigabyte GeForce GTX 1080 8GB, Power supply Corsair RM Series 850W, Windows 10 64 bit.

  • Thanks Antoine. Had missed that, but had been working on it for 3 hours, so my skills of observation had started to wane.

    When I finally extracted a 3 x 4 set of adjoining grids and loaded them into FS2 I found the tiles are not matching up. The reason would most likely be the missing textures on the right side of the tile. Tried files produced both with masks on and off with the same result. So appears to be a different but related issue to the one I described above. The screenshot below more obviously displays the problem just off the left wingtip. The plane was facing north so would fit with missing texture strip on the right

    Thanks, Chris

    Win 10 64-bit, 24GB RAM, i5-9400F @ 3.9, 6GB Nvidia RTX-2060

  • Hi Nick,

    yes good idea to allow Grid Square Selection with Level 12.

    It would also fine to activate Install scenery - even if I have to activate this manually after the conversion.

    (You could toggle the checkmarks, so this is the only allowed action ...)

    IPACS sometimes needs more time then someone thinks.

    Cheers, Thomas


  • Hello Nick,

    sorry the airports do not load, even in higher zoom.

    Hmm, frustrating about the airports as that's (I thought) quite a well tested feature. I'm also testing releases on a clean Virtual Machine and they work on there.

    The log isn't showing me much. If you really want to dig into this you could download SQLite Browser to open up the AeroScenery database. The table is FSCloudPortAirports.

    https://sqlitebrowser.org/

    Does anyone else see airports in 0.5 after clicking the "Show Airports" button and zooming in?

    Please note that "Open In Map" only works when a grid square is selected. I should be greying it out when nothing is selected.

    AeroScenery - Easily create photoreal scenery for Aerofly

  • I find the airports from Fscloudport load in AeroScenery without a problem. I did notice that the first time I tried to activate this, that the airport flag was nearly off the side of the screen and not correctly placed. As soon as I moved the map around a bit, the airport flag jumped to the right spot. Since then they have reliably appeared in the correct locations. The further you zoom in the more accurate is the placement of the flag. Click on the flag to display the Fscloudport details.

    aerofly.com/community/index.php?attachment/9631/

    Thanks, Chris

    Win 10 64-bit, 24GB RAM, i5-9400F @ 3.9, 6GB Nvidia RTX-2060

  • (...)

    When I finally extracted a 3 x 4 set of adjoining grids and loaded them into FS2 I found the tiles are not matching up. (...)

    Exactly, that's what I'd have expected from the previous pictures, when testing the Aeroscenery tile in AFS2.

    I think there's a miscalculation somewhere to investigate...

    Cheers

    Antoine

    Config : i7 6900K - 20MB currently set at 3.20GHz, Cooling Noctua NH-U14S, Motherboard ASUS Rampage V Extreme U3.1, RAM HyperX Savage Black Edition 16GB DDR4 3000 MHz, Graphic Card Gigabyte GeForce GTX 1080 8GB, Power supply Corsair RM Series 850W, Windows 10 64 bit.

  • I did some testing and comparison with the various files created by the manual FSET and automatic AeroScenery processes and after a few false starts I think I have an answer.

    Hi Chris,

    Thanks very much for the detailed test. I can see you spent some time on that. :)

    What I noticed is that as you give GeoConvert more resolution (by increasing the image zoom / detail level) the "overspill" gets less and less for a level 13 size tile.

    For a random level 13 size tile in France:

    At image detail level 17 I get six(!) masks

    At image detail level 19 I just get a small strip at the bottom.

    At image detail level 20 an even smaller strip

    Maybe by limiting size 13 & 14 to image detail level 18+ and tweaking the decimal rounding in the TMC file we can fix this.

    I'll have to look at this more fully later as I'm out of time now, but thanks again for the test.

    AeroScenery - Easily create photoreal scenery for Aerofly

  • Note that you have different pixel sizes in X and Y between FSET and Aeroscenery sources. The coordinates will vary depending on the settings in each tool, but the pixel size should be the same if both use the same raw image resolution

    Hi Antoine, I had to try a couple of extracts from FSET as the zoom level are differently described. I didn't understand the pixel size density and ignored it. All 3 produced a correctly imaged output file regardless of the input resolution, so this may not be an issue for the purposes of the comparison. The image resolutions available in FSET and AeroScenery don't match so any direct pixel comparison may not work.

    FSET has the following resolutions. There are more lower res level (higher numerically) but I never go higher than 3 or the image quality to too low to use.

    0 -> 0.5m/pix

    1 -> 1 m/pix

    2 -> 2 m/pix

    3 -> 4 m/pix

    AeroScenery has the following resolutions

    15 - 4.77 m/pix

    16 - 2.39 m/pix

    17 - 1.194 m/pix

    18 - 0.597 m/pix

    19 - 0.299 m/pix

    20 - 0.149

    I tried levels 3, 2 and 1 in FSET on the same level 13 grid area and got the following pixel steps:-

    . lvl 3 - [steps_per_pixel][7.15255737304688e-06 -5.36441802978516e-06]>

    . lvl 2 - [steps_per_pixel][2.86102294921875e-05 -2.14576721191406e-05]>

    . lvl 1 - [steps_per_pixel][1.43051147460938e-05 -1.07288360595703e-05]>

    I only run AeroScenery at one level for the same level 13 grid and got:-

    . lvl 17 - [steps_per_pixel][0.0000113993883132935 -0.0000100093409463844]>

    Thanks, Chris

    Win 10 64-bit, 24GB RAM, i5-9400F @ 3.9, 6GB Nvidia RTX-2060

  • Hi Nick,

    Ah yes, when a square is the slected, the browser opens a map. But always the same wrong area. I think it is a comma point problem (again). Please see this shot:

    However google needs decimal points!

    Regarding airports:

    The database seems pretty empty for me, though around Verona their are some FsCloudport airports:


    I also attach my database. Can it also be a locale problem with the coordinates?

    Cheers,

    Thomas

  • Hi Nick, I was wondering if there might need to be a link between the zoom level and the grid level size. In FSET this makes no difference to the outputted image other than the quality, but it uses a different method of extracting the data.

    One other random thought I had was perhaps there is an small inaccuracy in the information in the TFW file created by AeroScenery. The way I understand it the TFW files gives the coordinates of the top left point of the source image as it is represented on the planet. It then reports pixel step information to allow Geoconvert correctly map the rest of the image on the planet. The TMC file then comes along and gives GeoConvert top left and bottom right coordinates of the section to be extracted for the FS2 tile. If either the top left position or pixel step detail aren't accurate in the TFW file, then Geoconvert gets the wrong placement of the source image on the planet. It then extracts the section of the image it needs based on information in the TMC file. GeoConvert is then selecting the wrong section of the source image to create the FS2 tile. This would explain both the image overlap produced by the masks setting and the missing right hand section of the output image.

    Antoine pointed out that the pixel steps in the FSET extracts I produced didn't match (or even come close) to that used by AeroScenery (see my reply to Antoine above). Yet each of these pixel steps use in FSET allowed Geoconvert to correctly create data from the source image to the FS2 tile. Could this be the issue? Just a thought.

    Thanks, Chris

    Win 10 64-bit, 24GB RAM, i5-9400F @ 3.9, 6GB Nvidia RTX-2060