Agisoft Metashape
Agisoft Metashape => General => Topic started by: Dmitry Semyonov on January 03, 2012, 04:33:05 PM
-
Pre-release version of Agisoft PhotoScan 0.8.5 is available for download.
This pre-release is considered as unstable and is not recommended for production use.
Please let us know if you have any problems using this version.
Please post problems and/or questions concerning other PhotoScan version in separate topics.
This topic will be removed when the final 0.8.5 version is released.
Note
Project files created with 0.8.5 version can't be opened with earlier versions.
EDIT:
Updated to version 0.8.5 build 1376.
EDIT:
Updated to version 0.8.5 build 1398.
EDIT:
Updated to version 0.8.5 build 1419.
EDIT:
PhotoScan 0.8.5 build 1423 was finally released. Thank you for useful feedback.
Please use the links on the main PhotoScan pages for download:
Agisoft PhotoScan Standard edition (http://www.agisoft.ru/products/photoscan/standard/download)
Agisoft PhotoScan Professional edition (http://www.agisoft.ru/products/photoscan/professional/download)
-
After many tests with different data, my conclusion is that the Align Photos.. has serious problems. :(
Apparently it is very fast, but ineffective. In previous versions works much better. I hope you can solve. I think a lot of testing with the innovation they want to do. ;)
-
I also did some testing and I've got mixed results. In some cases, the results are a lot worst than in the latest versions. Some models that worked great in the 0.8.4 version get messed up in this version. In some cases all the model is crunched into an indistinguishable amalgam of points with the cameras spreaded allover, with nonsense positions. In a few cases, mainly related with terrain models from aerial photography, results are obtained a lot faster and seem to be better than in version 0.8.5. In one case, both the model and the orthophoto are more precise than in any previous version. I've tested some of the new features, which are all very interesting and useful. I had some problems in the Ground Control pallete. This is a great feature but, when I tried to check errors from the photos the model simply vanished. We need a new manual to some of these new features. I would like to see feature improvements in the Matches palette too. A manual edition of matching points could be a huge addition. This version is a very promising version.
-
Hello,
Thank you for reporting.
The problem with the alignment could be caused by the markers used. For now 0.8.5. uses them by default as correct matches and it may cause a misalignment if the markers are not placed precisely. On the datasets without markers the new alignment should work better then in previous versions.
Could you please inform us if the wrong alignments were in projects with markers?
-
Hi Alexey
All my tests were made without any markers. Only used the default alignment settings, without pair selection, markers or masks in the photos.
Cheers
-
Hi Alexey
I also do all my tests were made without any markers. :-\
Regards,
Diego
-
While using photos from UAV's or Kite Aerial Photography, I'm getting quicker and better results! Maybe it's an easy problem to fix. Most of the errors that I got came from very non favourable datasets. Aerial photography, with a good amount of superimposition, is giving me better results.
Cheers
-
From the 1224 version to 1330 version, the model often have the wrong texture case,I used the same data?Earlier version of the 794, there was no texture error, but the texture is very slow to generate :)
-
I have problems reading in ground control data for markers. No matter if I specify the markers in csv or the new xml file format version....
cheers
-
Hello wrumm,
Could you please provide the sample xml or csv file that has problems with import.
-
Hello wj,
Could you please provide screenshot that describes the problem?
-
Hi Alexey,
please find atteched the txt (tab delimited) and xml file.
The xml file was produced by exporting from the new groundcontrol panel and modifying some points manually in a textfile. During import nothing happens. The values remain the same.
Same thing with the txt file, but this one was created by hand, but I sticked to the manual.
cheers!
-
Problems with Ortophoto generation remain from version 0.8.4. If we try to render an ortophoto from a site with geometrically defined features mixed with terrains, like houses for instance, most of the times roofs or straight walls will get higly contorted.
There are also some problems with visualization, under macosx, when models exceed about 20 000 000 faces. Photoscan easely crashes beyond that limit.
When we start a batch with photo alignment and mesh generation, things will get done but, after that, the system gets very slow or completly blocked. It seems to be some memory management problem or memory dumping must be optimized. From activity monitor, we can see that Photoscan still blocks huge amounts of memory even with all the processes finished.
Cheers
-
Hello wrumm,
Thank you for the data provided.
Check if importing .txt file works correctly if quotes near the names of the markers are removed, because it should work in this case.
And .xml import problem will be fixed in the next update.
-
Hello RHenriques,
Thank you for the reporting. We'll check the memory problem.
Could you please also check if the visualization of large models is faster if VBO flag is unchecked in Preferences.
-
Hi Alexey Pasumansky
With the VBO flag unchecked, as you suggested, the model visualization is very fast. I could manipulate (rotate and zoom) a 40 000 000 faces model without any crash problem or any slowdown! What's exactly the fuctionallity of the VBO flag?
Thanks a lot.
Cheers
-
A Vertex Buffer Object (VBO) is an OpenGL feature that provides methods for uploading data (vertex, normal vector, color, etc.) to the video device for rendering.
-
Thanks Diego
Something must be wrong with the VBO feature then. It would be interesting to simplify the visualization but, after Alexey's suggestion to turn it off, I can see that it can be problematic under Macosx. Maybe it as something to do with ATI's implementation of this feature in the Macosx drives.
Cheers
-
Maybe it has to do with ATI. For Nvidia is essential to have it enabled to go fast visualization.
-
For the exact same scene and set of photos (masked) 0.8.4 aligns them perfectly while 0.8.5 fails...
F/
-
I've been having trouble exporting 3D PDF's. The model will export - but if I try to export a texture the PDF is blank.
-
Hello ChetWalker,
What is the size of texture atlas you are exporting?
-
Any eta on when the 0.8.5 is gonna be released?
-
Any eta on when the 0.8.5 is gonna be released?
Just in case you...
0.8.5 has been out since January 3rd.... so you want to say 0.8.6....
f/
-
Just in case you...
0.8.5 has been out since January 3rd.... so you want to say 0.8.6....
f/
Just in case you...
0.8.5 BETA has been out since January 3rd... so I wouldn't say 0.8.6....
Read the title of the thread. Does it say post-release?
;)
-
ok..
f/ :)
-
Hi guys,
Playing with the 0.8.5 pre-release, and I seem to longer be able to import a mask for the image via a photo's alpha channel.. I wrote a batch processor this end to detect blue/green screen backdrops that exports a TIFF image with alpha that acts as the masking information for PhotoScan. This has been working properly till this pre-release.
Unfortunately the console does not give me any feedback except to its finished
cheers
Mike
-
Hi guys,
Playing with the 0.8.5 pre-release, and I seem to longer be able to import a mask for the image via a photo's alpha channel.. I wrote a batch processor this end to detect blue/green screen backdrops that exports a TIFF image with alpha that acts as the masking information for PhotoScan. This has been working properly till this pre-release.
Unfortunately the console does not give me any feedback except to its finished
cheers
Mike
look here...
http://www.agisoft.ru/forum/index.php?topic=331.0
solution for now.. import your masks using 0.8.4, save it, then open it in 0.8.5.. voila. :)
f/
-
have already been doing exactly as you mentioned.. strange I didn't see your post on this bug before.
cheers
-
Hi everybody,
I am currently working through version 0.8.5 build 1376 on UBUNTU 10.04.3 AMD64 and found that a couple of formats have been removed from the format selection drop down box on the export-points-dialog.
There are only 4 left: OBJ, PLY, XYZ and ASPRS LAS, which is new according to changelog
In 0.8.4 we have:
PS-Structure, OBJ, PLY, CHAN, BOUJOU, Bundler, OMEGAPHIKAPPA
Is this a bug? I hope so, because I need the other export formats for the pointcloud urgently.
If this is not a bug, but a "feature", I really would appreciate to have the same export formats as in version 0.8.4.
Cheers
ju523m
-
Hello Thomas,
Several export points formats that include or contain only camera positions were moved to Export Cameras dialog in Tools menu.
-
Hi Alexey,
oops, I did not see that, sorry.
Thanks, I am happy again.
cheers
-
Hello... where can I DL the last 0.8.5 pre-release...?
Thank you
-
Hello FoodMan,
All links in the first message (http://www.agisoft.ru/forum/index.php?topic=322.msg1291#msg1291) of this topic are updated. The latest 0.8.5 build available is 1376.
Please inform us if there are problems with any package.
-
Thanks a lot Alexey... got it.. 8)
f/
-
I have just tested new version.
- Aligning images produces much worse results. Instead of one scene I get several unrelated fragments.
- Geometry reconstruction is much quicker but not so good as in earlier versions
- Files generated with new version cannot be open with earlier versions
- I really like option to select part of mesh/cloud and to see what images are part of it :)
-
Hello Dino,
Could you provide us sample data set that is always aligned good in 0.8.4 and fails to align in 0.8.5?
And what parameters for geometry reconstruction should be used to see the difference in resulting quality?
-
hello... just a little note Alexey..
with the last build (0.8.5 - 1376) I have output an OBJ file... for 3.5 mil polys, the file size was a bit more than a GB.. I opened the file in modo, resaved it in obj. format.. only 330 mb...
any view on this...?
f/
-
Hello FoodMan,
Could you please check if there is the same number of decimal digits in both files?
-
Hi all,
i have a question concerning 0.8.5 1376, Ubuntu Linux 10.04.3:
Where has function "Propagate mask" from Photo menu gone?
Thanks for your help
kind regards
-
Hello Thomas,
The result of Propagate mask feature can be achieved in PhotoScan 0.8.5 by exporting mask of the single photo and importing it back for the another photo/chunk/workspace. In this case mask filename should not depend on filename.
-
This latest beta is crashing everytime I leave a model building geometry, under MacOSX. I've sent a crash log after one of these crashes.
Cheers
-
Same on my iMac.
-
Thanks for reporting.
We'll check it.
-
I was able to avoid some of these crashes by turning off the "Enable VBO support" option.
Cheers
"Edited"
After testing with a heavy 3d terrain model, I can confirm that disabling the "Enable VBO support" option in the preferences will stop the crashes under MacOSX.
Cheers
-
After testing with a heavy 3d terrain model, I can confirm that disabling the "Enable VBO support" option in the preferences will stop the crashes under MacOSX.
Sorry to disappoint you but I have to confirm the opposite. Disabling VBO support does NOT stop the crashes.
What I also had once was a successful termination of the "Building geometry" process without the creation of a 3D-model. It just ended the process.
-
Hi Oli63
In my case the crashes are gone, at least in my 2 latest tests with VBO off. This includes the creation and further visualization of 3d models. Considering that are models that take around 7 to 10 hours to finish I could not make more tests.
Since we are having mixed results, can it have something to do with different graphical GPU's.
Cheers
-
Hi all,
From our last UAV project, processed with 0.8.5 beta I can confirm that the photo alignment does not work as expected. Using the GPS positions as initial values produced a lot (>50%) of unaligned photos. The generic matching algorithm matched all photos.
The final DEM was lower in quality than what we normally expect, but this may also be partly due to the type of area and illumination conditions (very low sun).
So until you have fully debugged the new algorithms, I'd prefer the option to also use the old algorithms for photo matching/DSM construction when you release 0.8.5.
Otherwise, I like the new features being implemented.
Some thoughts about projection systems: it seems as if they're 'hard coded' into the exe. Maybe useful if you could put them in a text/xml file that we can edit so that we can include our own projection systems. For example, the default Belgian projection system has wrong parameters in the Proj4/GDAL library...
Best regards,
Jan
-
Hello Jan,
Could you please specify which build of 0.8.5 you were testing?
And regarding projection system - what's its EPSG code?
-
Hi Alexey,
Thanks for the follow up.
It was build 1398 of March 1st. Meanwhile I can confirm that on the same set of images, v0.8.4 has no problems in photo alignment.
The projection system I was referring to is the Belgian projection 31370. It's definitely in your list, but if derived from GDAL the towgs84 parameters are often wrong. Instead of:
+towgs84=106.869,-52.2978,103.724,-0.33657,0.456955,-1.84218
they should be
+towgs84=-106.869,52.2978,-103.724,0.33657,-0.456955,1.84218
Jan
Hello Jan,
Could you please specify which build of 0.8.5 you were testing?
And regarding projection system - what's its EPSG code?
-
Hello Jan,
We've checked the coordinate conversion from Belge 1972 to Belgian Lambert 1972 (Datum Reseau National Belge 1972) and the result was the same as in free coordinate converters available on the web.
Could you please tell us how is it possible to reproduce or detect the mistake, since PhotoScan doesn't use GDAL?
-
Hi Alexey,
Thanks for the reply. I think you are right and it's better indeed to leave it as is. We actually raraly/never use Belgian datum for geographic degrees. It's either lat/lon in WGS84 (GPS positions) or Belgian datum as projected (Lambert). There's plenty of tools around that do this conversion for us so for us there's no need to include datum conversions in PhotoScan reprojections.
Best regards,
Jan
Hello Jan,
We've checked the coordinate conversion from Belge 1972 to Belgian Lambert 1972 (Datum Reseau National Belge 1972) and the result was the same as in free coordinate converters available on the web.
Could you please tell us how is it possible to reproduce or detect the mistake, since PhotoScan doesn't use GDAL?
-
Hi!
Thank you for the very useful new features in v0.8.5!
In our case we youse a metric medium format camera. The lens distortion is highly accurate estimated in a lab calibration, but the camera constant is not very stable. Hence it would be very useful if every parameter of the camera calibration could separately be enabled/disabled in the Optimize Photo Alignment-Tool (not only aspect, skew, tangential distortion).
Another helpful feature would be a precision information of the estimated camera parameters (and for the other estimated parameters of exterior orientation too...). Based on this precision information we could decide whether a camera parameter is significant different from zero or not, and then it could be enabled/disabled in the further adjustment.
Best regards
Fabian
-
Thanks for the new features in the new build Alexey and team, and for your hard work on delivering consistent updates.
-
Hey guys
Which graphic card do you recomend to be able to get stereo vision?
-
Hi,
I use Nvidia Quadro 5000.
Regards,
Diego
-
Hi there. I finally got a license (student). I downloaded the lastest version yesterday and one of the most hated issues with photoscan was gone, wich was the loooooong time it did take in the previous demos to match the photos and get camera positions.
Now I would love to have the feature of saving the "dense point cloud" that I believe is mad but not rendered prior to meshing, as some times you may place 2000000 polygons but half is used for non desired geometry or other areas so have to recompute again. (this is even more painfull in themo as you cant save previous stages).
-
Hello Paulov,
You can set "Face count" parameter to "0" in Build Geometry dialog and perform decimation later after all the unwanted model parts are removed. Zero value means that mesh would be have maximum number of faces for the current reconstruction parameters.