Agisoft Metashape
Agisoft Metashape => General => Topic started by: Dmitry Semyonov on January 30, 2013, 12:56:28 PM
-
Pre-release version of Agisoft PhotoScan 0.9.1 is available for download.
This pre-release is considered as unstable and is not recommended for production use.
Please use the links below for download:
Agisoft PhotoScan Standard edition
Windows 64 bit (http://downloads.agisoft.ru/win/photoscan_0_9_1_x64.msi)
Windows 32 bit (http://downloads.agisoft.ru/win/photoscan_0_9_1_x86.msi)
Mac OS X (http://downloads.agisoft.ru/mac/photoscan_0_9_1.dmg)
Linux 64 bit (http://downloads.agisoft.ru/linux/photoscan_0_9_1_amd64.tar.gz)
Agisoft PhotoScan Professional edition
Windows 64 bit (http://downloads.agisoft.ru/win/photoscan-pro_0_9_1_x64.msi)
Windows 32 bit (http://downloads.agisoft.ru/win/photoscan-pro_0_9_1_x86.msi)
Mac OS X (http://downloads.agisoft.ru/mac/photoscan-pro_0_9_1.dmg)
Linux 64 bit (http://downloads.agisoft.ru/linux/photoscan-pro_0_9_1_amd64.tar.gz)
PhotoScan Professional 0.9.1 User Manual (http://downloads.agisoft.ru/pdf/photoscan-pro_0_9_1_en.pdf)
PhotoScan 0.9.1 Change Log (http://downloads.agisoft.ru/pdf/photoscan_changelog.pdf)
PhotoScan 0.9.1 Python API (http://downloads.agisoft.ru/pdf/photoscan_python_api_0_9_1.pdf)
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 0.9.1 version is released.
Note
Project files created with 0.9.1 version can't be opened with earlier versions.
EDIT:
PhotoScan 0.9.1 build 1703 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)
-
Hello Alexey,
I looked at the change log and didn't see anything about depth maps. Does 0.9.1 solve the depth map problem for projects in which camera positions are within view of other cameras (http://www.agisoft.ru/forum/index.php?topic=971.0 (http://www.agisoft.ru/forum/index.php?topic=971.0))?
-
Hello RalfH,
There are no changes for skipping pairs during depth maps calculation. Unfortunately, this task is not on top of our "to do list", since such camera configuration is not usually used and in most cases is not optimal.
-
Hello Alexey,
I know that this is not an optimal camera configuration, but in some cases it is the simplest (and sometimes the only) way to work (i.e. when access is restricted or when there is very little time). In the case of one of my projects, walking on the surface was leaving footprints, so I could only walk across the feature once and take pictures in the direction in which I was walking. I think it would be good to make PhotoScan even more flexible for "unusual" configurations.
-
hello Alexey..
If I can ask... what exactly do you mean by
Added full utilization of HDR depth
thank you
f/
-
Hello Foodman,
This means that no conversions to 8-bit per channel are performed during any processing step.
-
very cool.... do I guess right to say that Pscan is now using all the extra information from 32 bits images... 8) 8)
f/
-
Why does the above link not point to the 0.9.1 log file but only the Version 0.9.0 build? Where could I check the latest improvements?
CU,
Geert
-
This pre-release is considered as unstable and is not recommended for production use.
[...]
Note
Project files created with 0.9.1 version can't be opened with earlier versions.
This means that it is very difficult to test 0.9.1 and also keep working with 0.9.0 without having to repeatedly uninstall/install the two different versions.
-
Hello Geert,
Please update browser window if checking pdf online.
-
We calculate the time depth is about 30-40 percent better.
470 product images are calculated in 1h.40 min.
instead of 3 hours. !
revelation :)
-
The improvements in the new release and arbitary/mosaic mode are incredible :D
Great job Agisoft!
;D
-
Hello 2 all,
also unbelievable improvements in building geometry.
70 pictures on "HIGH" before: 4845s now:2890s
GPU: 7970 ghz edition
Well done, congratulations!
Best regards
EDIT: second attachment results with 9 pictures on Ultra
-
The texturing is sooo improved... amazing !!!!! Great great Job Guys !! 8) 8) 8) 8)
f/
-
hello,
one thing I have to mention is the optimizing step. Only to check I optimized an old project with the new version of photoscan. I didnt change anything in the settings and wanted to compare the results.
Scales were not used, K4 disortion disabled. For optimizing were used only GCP with an accuracy of +-2cm (marker accuracy set to 0 in settings). With the version 0.9.0 there are no probnlems. After
two optimizing steps (and gradual selection) the pointcloud had already a good quality (error under 10cm). With the pre-release version there were no way to fit the pointcloud under 20 cm, even after lot of more optimizing steps. Please note that the settings were the same. What I noticed was, that the optimizing step in pre-release works with less number of iteration?? That is what I saw in the "console". The weights are not growing after each iteration. Maybe it happens in the backround now. I dont know.
After this I tried to handle the problem with the new scale bars and the K4 fit, but there was no improvement to get the bundle adjustment under 20 cm. In fact in my opinion scale bars wouldnt improve the result when using GCP. Because the information about the distance between the GCP already exists in the coordinates.
Any idea how to fix this? Is it a known problem??
Thanks a lot!
-
Can't believe the speed improvements in the new version guys !
Well done Alexey and the team! 8)
I have just tested with a set of photos from one of my LED dome prototype sessions (more than a year ago) The whole build used to take 1hr 46 mins...now it takes 36 mins!!!
I use quite an old MacPro 3.2 8-core...and Agisoft PS doesn't seem to like my graphics card, So this new huge speed improvement is very welcome.
Also the new texturing blending seems amazing..what would normally take me a lot of work to fix in photoshop is now nearly perfect straight out of Photoscan.
Keep up the good work!!
-
Can't believe the speed improvements in the new version guys !
Well done Alexey and the team! 8)
I have just tested with a set of photos from one of my LED dome prototype sessions (more than a year ago) The whole build used to take 1hr 46 mins...now it takes 36 mins!!!
I use quite an old MacPro 3.2 8-core...and Agisoft PS doesn't seem to like my graphics card, So this new huge speed improvement is very welcome.
Also the new texturing blending seems amazing..what would normally take me a lot of work to fix in photoshop is now nearly perfect straight out of Photoscan.
Keep up the good work!!
Yes indeedy! these guys have made some incredible improvements in this latest release and I have a feeling some more are on the way! :)
-
Great improves.... texture is the best!
-
My first impressions after some testing are mostly positive: improved mosaic texture, improved model reconstruction for large flat areas. One thing I noticed is that occasionally my PC froze during processing even though memory usage was only approx. 50% (even the status update of the task manager was interrupted for a few minutes); that's something I hadn't noticed with PhotoScan after version 0.8.
-
Something is wrong in this version. While trying to reconstruct a model from about 86 cameras, everything is normal with the point cloud generation. After that, while trying to reconstruct the model, no matter what parameters are used, a message with "Can't build model" appears. The same photographs were used without any problem in the previous version. Images are .CR2 files from a Canon. I'll try with other models in JPEG format to see if there's something connected to file format or to my machine.
-
Hello, I have been using the demo version for a while now, I want to purchase a standard license. Will I be able to recieve all the updates that Agisoft Guys provide for the standard software in the future? Thank you!
-
After my previous message I did some tests and the "Cant't build model" problem seems to be linked to the NVIDIA GeForce 9600M GT graphical card. If I go to the OpenCl in the preferences and deactivate it, the model is built without problems. Later I'll copy the console text that appears after the failure.
Cheers
-
Koyko, yes you get free updates.....
-
From my tests so far, I've noticed that the 3d models from aerial imagery tend to be worst (much more triangulated), in version 091 than in version 090 (images attached), using the same parameters. The calculation must be different because the final number of polygons is very low for the same "0" value in the number of desired polygons. It seems that, by default, Photoscan optimizes or decimates the model instead of produce the maximum possible number of polygons. How do we use the same functionality of the last version to produce the biggest amount of polygons?
By the contrary, orthophotos are a lot lot better in 091 that in 090. Sometimes this difference is impressive!!
Considering the "Can't build model error", these are the Graphical Card specifications:
OpenGL Vendor: NVIDIA Corporation
OpenGL Renderer: NVIDIA GeForce 9600M GT OpenGL Engine
OpenGL Version: 2.1 NVIDIA-8.0.61
Maximum Texture Size: 8192
Quad Buffered Stereo: not enabled
ARB_vertex_buffer_object: supported
ARB_texture_non_power_of_two: supported
loaded project in 4.73794 sec
Using device: GeForce 9600M GT, 4 compute units, 512 MB global memory
max work group size 512
max work item sizes [512, 512, 64]
max mem alloc size 128 MB
And this is the error:
stereo.cpp line 203: error CL_BUILD_PROGRAM_FAILURE
Finished processing in 0.157732 sec (exit code 0)
>>>
PS: I've just tested a new run and, if I put 10 000 000 faces in the "Face count" option the same 9 plus million faces, as in the previous version, are generated. It would be great to get these biggest possible amount using "0" in this option, as in the previous version or something a little more intuitive. In the case of this model I knew the maximum amount of faces in advance but, in a new project, the only option will be using a very big number to avoid guessing....
-
RHenriques,
I had a similar problem and Alexey told me that since 0.9.0, the behaviour of PhotoScan with regards to the desired number of faces has been changed: "zero" doen't mean unlimited anymore, so if you don't want the model to be decimated you should put a very high number. Personally, I preferred using "zero" for unlimited as well...
Alexey: perhaps it would be possible to add a tick box where the user can choose "no decimation"?
-
Hi RalfH
RalfH said "Alexey: perhaps it would be possible to add a tick box where the user can choose "no decimation"?"
I agree with RalfH. The suggested option for "No decimation" it's a better solution.
Other than this, and the "Can't build model" problem that I'm experiencing, this version is a very good and stable update, mainly concerning the calculation speed, the quality of the texture and the improvements in the orthophoto generation.
Cheers
-
this version is a very good and stable update, mainly concerning the calculation speed, the quality of the texture and the improvements in the orthophoto generation.
Very much this. I can confirm that I have had 5 powerful PC's running, processing 100's of scans reliably for 7 days straight in the last month. Not one crash. You can leave it running over night without fear.
Super stable.
-
Increased speed, better texture. but this only works on individual chunks.
calculations connected by chunks, ending the outbreak of strange artifacts.
a very serious problem.
I do not recommend you use this version if you have an extensive project 300 photos or affect more.
http://www.agisoft.ru/forum/index.php?topic=992.0
-
I want to underline what has already been said about the new 0.9.1 version. I tested one of my models consisting of 109 cameras, showing a landscape with lots of buildings on it. The result is a fantastic improvement. A very much more precise mesh and a texture without all the distortions of the same model calculated in 0.9.0. What excited me most is that the textures on vertical walls are veeeery much better now. Really impressive, great work! Thank you Dmitry.
-
Increased speed, better texture. but this only works on individual chunks.
calculations connected by chunks, ending the outbreak of strange artifacts.
a very serious problem.
I do not recommend you use this version if you have an extensive project 300 photos or affect more.
http://www.agisoft.ru/forum/index.php?topic=992.0
Best to talk directly with the Dev team, share data etc so they can test that issue you are facing.
-
Hello RHenriques,
And this is the error:
stereo.cpp line 203: error CL_BUILD_PROGRAM_FAILURE
Thank you for reporting.
Are you using Mac OS? If yes, than the problem will be fixed in the next 0.9.1 update.
-
Hello RHenriques,
And this is the error:
stereo.cpp line 203: error CL_BUILD_PROGRAM_FAILURE
Thank you for reporting.
Are you using Mac OS? If yes, than the problem will be fixed in the next 0.9.1 update.
Hi Alexey
Yes, I'm using MacOSX. Despite this small bug, that I'm sure you'll correct very quickly, this update is very robust so far. I'm so impressed with the improvements in the Orthophoto generation that I guess that I'll have to repeat several projects made with the previous versions!!! And this can only be a good thing for current and new PhotoScan users.
Cheers
-
In Version 0.9.0 it was possible to export a 3d PDF in a geographic coordinate system. This option is now disabled in version 0.9.1 (always export in local coordinate system).
I used to activate the 3d measurement tool of the exported PDF (enable comment function for reader) in adobe acrobat pro. Afterwards it was possible to measure 3D coordinates in the geographic coordinate system within adobe reader - a very impressive function for our costumers.
It would be very nice if it were possible to choose a coordinate system in the PDF export dialog, and also a tick to activate directly the pdf comment function to enable 3d measurements.
Cheers
-
Great improvements in this new version. Just couple of minor bugs... cant seem to check/uncheck all selected cameras/markers. Sometime works, sometimes doesnt. Also, how does the create scale bar works?
cheers,
Javier
-
Hello Javier,
Could you please specify OS you are using so that we'll try to reproduce the issue.
Scale-bars can be created as a distance between two markers. So at first you need to place two markers, then switch to Create Scale Bar tool and point to markers consequently - this operation will create a scale bar instance in the Ground Control pane.
If you wish to use scale bars in optimization process or for model scaling you can input known distance as Source value in the Ground control pane and check (turn on) the scale bar.
Also scale bars could be used for linear measuring purposes - estimated distance will be shown on the corresponding Ground Control pane tab.
-
Thanks Alexey. I'm working under Windows 7. Cheking/unchecking of selected markers seems to work only after I have manually checked each marker.
As for the scale bars, when I add one and check it... the estimated distance actually appears under the error (m) heading which makes it a bit counter intuitive, particularly if I use it in the optimization process as then it changes to the actual error.
-
Hello Javier,
Checking all selected cameras works only for those of them which have source coordinates. While unchecking should work in any cases (CTRL+A -> Right click -> Uncheck).
As for the scale bars that will be used for measuring (and do not have any source values) should not be checked, since in this case their distance will be automatically assigned to zero. To check estimated value you need to switch Ground Control pane to Estimated values view.
-
Alexey, im get problem with selecting images -im have pressed the CTRL key and clicking on images that want disable, but afther not same period of time the selection is lost, and im must start from begining to select images. mostly can set upto 3-10 images then it lost selection. have tested to change Keyeboard, mouse, try on another 2 PCs but same problem. And its not just this version... Haveo thers same problem too ??
-
Hello Wishgranter,
Do these problems appear in Photos pane? If yes, then it is connected to the mouse-click during the mouse movement. In this case PhotoScan starts multiple selection and resets the previously selected files.
-
While we're at it... it would be awesome to have an images order possibility.. like alphabetic , non-aligned, disabled, etc... and reverse... so this way a simple Shift-click... and voila.. :)
f/
-
I'm glad to see improvments in the production of orthoimagery. using 0.9.1 build 1632 64-bit I revisited an old project to see if there were improvements in the orthomosaic function. Without reprocessing geometry or camera alignment I generated a new orthoimage. I had hoped the bug preventing masked regions from being output in mosaic mode would have been fixed. Doesn't look like the case, but it does appear that there is a fair amount of color correction taking place. These two captures show the differences in color corrective output between v 0.9.0 (1584) and v 0.9.1 (1632), it seems like there was heavy balancing done on the blue channel. This is nice and all but the masking bug really limits the aesthetic quality of the orthomosaic.
Also, i have been running into issues with limitations in VRAM of my Quadro 5000. Im running a dell Precision T7600 with dual octocore xeon E5-2687W cpus at 3.1ghz, 128gb ECC DDR3, 8 TB RAID. I usually have no issues with processing of large image sets but i run into performance problems trying to display meshed models consisting over more than 40 million faces. It appears that the VRAM fills up and brings the machine to its knees. Is there any way of more efficiently displaying large meshes or should i considered trading this card in for a Geforce Titan 6GB?
(http://imageshack.us/a/img21/1048/v90.png)
(http://imageshack.us/a/img805/5429/v91.png)
-
Hello, tha Vram issue is anyoing.. yes it will help, every 1 GB of ram is aprox 45 Mil triangles, without texture...... But realisticaly, going over 100 Mil will be problematic even then ( its something that was unthinkable few years back )
So teoreticaly, without texture, it should be possible to open and work with 200 Mil mesh ( with some reserve 250 Mil )
Im working with 7970 3 GB VRAM, an its get problem to go over 100 Mil ( propably a driver issue, it seems that it could not get properly in VRAM ) Im set decimation to 50 Mil, and its quick in viewport
How much VRAM you have now ?
Can report us then how it work with TITAN ? :-)
-
I currently have two Quadro 5000 cards that came installed with the machine. Each has 2gb vram, non-SLI mode. To my understanding SLI wouldn't help since each card would load the same data. When reconstructing depth each seems to be fully utilized as individual computer clusters. Even at 40mil faces it is difficult to work with. If i can get my hands on a Titan ill definitely post some benchmarks. I use cloudcompare to work with the pointclouds, and have tried i-site studio 4 as well but both seem to hit the VRAM wall with large datasets...
Anyone know any good pointcloud analysis software that uses octrees for visual display?
-
The 64 bit Windows version seems to be broken. The installer won't launch at all for me.
Windows 7 x64
-
Hello Maharg,
Thank you for reporting.
We have updated installation package for PhotoScan Standard Win-64bit.
-
Beautiful. Thank you!
-
Export Orthophoto/Blending Mode Average crashes Photoscan.
I attempted to generate both tiff and jpg orthos with blending mode average. Both times Photoscan crashed. I successfully exported a mosaic with the same dimensions, so the problem appears to be specific to the pixel-averaged image export.
crash reports were sent (according to the reporter). Let me know if I can do anything else.
-
Mosaic issue?
I generated two orthophoto mosaics with 0.9.1 and was surprised to see some artifacts in the mosaicking. The surface model still looks OK, but there appears to be least one horizontal seam in the image - right in the middle. It's not completely clear to me if it is continuous or not, but I did find it exactly in the middle of two different images I generated from two different areas of a model.
Projection was NAD83 UTM 10. orthomosaic bounds were specified in integer meters.
-
Hello andyroo,
Please update to the latest build (1641). Hopefully both issues are fixed there.
-
This means that no conversions to 8-bit per channel are performed during any processing step.
Hi Alexey,
So if photos are import as half-precision floating point .exr files will all calculations be floating point operations? What about if photos are import from 16-bit integer tifs? Are the values converted to floats so that floating point operations can be used or are calculations done as integer operations.
I'm interested in this in the context of avoiding quantization problems in a linear scene-referred pipeline, i.e. floating point operations are good integer operations, not so much.
Thanks,
-
Hello Andy,
Currently single orthophoto file in TIFF format is limited by 2 GB. If it is not possible to generate smaller file using the current settings (export resolution) "Not enough memory" message will be displayed during the export orthophoto.
-
Hello jedfrechette,
Both 16-bit integer and half-precision floating point data will be converted to floating point.
-
Hello everybody,
I'm playing with this pre-released. But I'm having a problem to scale the model. I checked the tutorial.
I set two markers, but I can't find 'set reference scale'. However I can create scale bar. How could I set the measure?
thankssss
-
Hello b_moler,
In 0.9.1 scale bars could be now used both for reference distance assignment and for measuring.
If you wish to use the scale bar as reference distance (could be several in a single chunk) you need to place two markers, then switch to Create Scale bar tool and point consequently to the both markers pointing to the scale bar ends.
After that the reference distance should be input manually in the Source tab of the Ground Control pane for the corresponding scale bar. Do not forger to check on the scale bar.
In case you are using scale bars for measuring the only difference with the procedure described above is that you need to uncheck the scale bar in the Ground control pane, while the estimated distance will be shown for such scale bar in the Estimated tab.
-
Currently single orthophoto file in TIFF format is limited by 2 GB. If it is not possible to generate smaller file using the current settings (export resolution) "Not enough memory" message will be displayed during the export orthophoto.
I was able to generate mosaic orthophotos with exactly the same dimensions as the RGB average orthos that gave me an out-of-memory error. (retracted - I think I mistyped the bounding boxes. Sorry I should have tried to duplicate the error first)
-
Hello Andy,
Could you please provide the log file for successful export and for export with memory problem?
And just for case please also check you are using 64-bit version and not 32-bit.
-
Hello Andy,
Could you please provide the log file for successful export and for export with memory problem?
And just for case please also check you are using 64-bit version and not 32-bit.
I generated about a dozen new averaged images and have not been able to duplicate the problem. I think I must have mistyped the bounding coordinates. sorry for the false report.
Right after I wrote the above, I managed to crash Photoscan while exporting a RGB average image. The crash also occurred immediately after relaunching Photoscan and trying to perform the same export, and when I switched to exporting a mosaic for the same chunk (instead of RGB average), so this may be a different problem. An error report was generated for each crash, but I am attaching the logfile (as of the first crash) to this reply anyway.I had started a new logfile immediately before beginning to export these orthoimages.
This is reproducible with this project file, at least on my machine). I can upload the project file (and imagery if needed) for the whole project or just this chunk if it would help.
(also tried with build 1647, and it still crashes. crash reporter appears to have sent the dump)
Andy
-
Not sure where to suggest features, but this FaceWorks demo by nVidia could potentially improve huge mesh rendering.
http://youtu.be/5d1ZOYU4gpo
-
Not sure where to suggest features, but this FaceWorks demo by nVidia could potentially improve huge mesh rendering.
There's a features section on the forum.. "Feature Requests" http://www.agisoft.ru/forum/index.php?board=11.0
That video above uses a very sophisticated and expensive Lighstage to capture data, most probably using Phantom v-Series cameras, also very expensive (average of $250,000 each)
-
Photoscan Pro 0.9.1 build 1647 seems to have a problem aligning uncompressed tiffs. I converted eleven Canon RAW images to uncompressed tifs in Photoshop CS6. Photoscan Pro 0.9.0 (1586 64-bit) aligns perfectly. In the beta, the model aligns but it is almost impossible to view. It is as if you are a few millimeters from the model.
I can send you my tifs if you want to recreate the problem.
-Mark
-
Export depth seems to be broken in 1647. Also, it would be nice to be able to export all of the selected depth images instead of having to do them individually.
G
-
Hello Maharg and Mark,
Thank you for reporting. We'll check this issues.
-
Im get problems when im try export undisorted photos.. will prepare the dataset.....
-
Hello Mark,
We have checked your datasets and found that the cause of the problem. The images are geotagged with the identical GPS coordinates and PhotoScan 0.9.1 loads the EXIF information automatically on image load.
If you unreference the chunk (use Reset Chunk Transform option from context menu available in Workspace pane) the visualization problem will be gone.
-
after creating the mesh, I noticed the presence of these forms invented by the software .. someone has already happened?
-
Hello kiteman,
difficult to say without more information (e.g., images, 3D reconstruction settings, presence of vegetation). My first thought would be that some of the images are misaligned. Changing camera calibration settings from groups to separate and re-aligning the images can do wonders in that case.
-
Many thanks now i try calibration separate
There aren't tree
-
litle faster ??? aprox 10 % ??
im get issues with the mask generation from alpha, something weird, if images are in 2 folders, then just from one folder are mask generated..... will upload the dataset eventually ?? even afther moving to sam folder it could not create it..... someone same problem ?? And not every mask was generated from ALPHA...... the newest beta was in work..
-
Hello Wishgranter,
Please send us the minimal dataset that we can use to reproduce the problem.
-
another issue, when set to BATCH, set to create geometry and im set my limit on triangles. if im set any size say 25mil it set it to 200.000 triangles..... have set few other number but same result..... on other dataset too....
-
I have a similar problem to Wishgranter. When Batch mode is used Build Geometry always resets to Smooth and 200000 maximum faces and ignores what I set.
-
Do you mean that the values are not saved in the dialog box?
-
yes Alexey, it save not our settings in batch.....
-
Thanks for reporting, we've reproduced this problem and will fix it in the next update.
-
Hello Alexey,
Version 0.9.1 build 1659 not fixed calibration. Optimize photo alignment change the calibration parameters in ajusted.
Version 0.9.1 build 1654 is OK
-
Hello Diego,
Version 0.9.1 build 1659 not fixed calibration. Optimize photo alignment change the calibration parameters in ajusted.
Version 0.9.1 build 1654 is OK
Fixed calibration flag makes PhotoScan to use fixed values from Initial tab. In case you have different values in the Adjusted tab and wish to keep them during the optimization you can just uncheck all the parameters in the Optimize camera alignment dialog.
-
Fixed calibration flag makes PhotoScan to use fixed values from Initial tab. In case you have different values in the Adjusted tab and wish to keep them during the optimization you can just uncheck all the parameters in the Optimize camera alignment dialog.
Hello Alexey,
I understand, but I fixed calibration from the beginning and in earlier versions if fixed and did not change when It optimized parameters. But in this version, when fixed to the calibration from the beginning, the software changes the parameters when optimized, so this fixed. I think it's a bug.
Version 0.9.1 build 1659 not fixed calibration.
Version 0.9.1 build 1654 is OK.
Kind Regards,
Diego
-
Ciao!
codecs target, are useful only for GCP, or are used by the software as well as help in the process of alignment of the images?
Stefano
-
Hello Diego,
Could you please check that the problem is reproducible and describe the workflow, so we could check it on our side?
Hello Stefano,
Markers based on coded targets work similar to common markers - their could be used as valid matches when "Align selected cameras" option is used.
-
Many thanks
-
Hello Alexey,
does the new support for K4 radial distortion mean that fisheye lenses could be used?
-
Hello Oliver,
K4 could be adjusted only during the optimization step, however, if the cameras could be aligned the use of this parameter could produce better approximation and reduce the reprojection error.
But in case fish-eye images can not pass Align Photos stage successfully K4 could not be used to help anyhow.
-
Hi All,
I'm using 0.9.1 with great results.....however...
I'm finding that when I import my mesh into Mudbox is has quite a few ngon and tshapes etc. in the exported mesh. It means we can't use any retopology tools. I'm decimating the mesh down to 1,000,000 polys in photoscan BTW in case that may be doing it? I'm just assuming this is the issue?
Any ideas what I can do to get around this? Can I leave it RAW capture resolution ( we used to have the poly count as zero to get this)
Or any ideas how I can make the mesh useable?
Many thanks in advance.
-
Hello Subunderground,
PhotoScan operates only with triangle faces and exported mesh should contain only triangle as well. Could you please check that other polygon types were not created during the import?
I can also recommend to use Fix Topology option in View Mesh Statistics dialog (available from Tools menu) to remove topology issues (like duplicate faces, zero faces, multiple edges and etc.).
-
Alexey, please read email, im have some screenshots of the problem with sample mesh and little explanation.
-
Hi Alexey, thanks for getting back so quickly.
I've just tried the Fix Topology which has the same issue, and I'm now going through every export option to try and get one that works. I can also email you some screenshots if that would help.
Many thanks!
-
Hello Wishgranter,
Thank you, will check it.
Hello Subunderground,
The screenshots may help, also if you could share the project file and/or exported model it can be also helpful. If they are not so big please send the data to support@agisoft.ru
-
Hello Diego,
Please try build 1667, now Fixed calibration mode should be working properly.
-
Hello Wishgranter and Subunderground,
We hope that updated version will also fix the topology problems and other issues with faces.
-
Alexey,
I just tried build 1667 with a project with fixed cameras and errors jump from ~0.25 pixels per photo to over 100 pixels.
The project actually has four cameras - two have to be held fixed and the other two can, need, to be calibrated. The alignment went great with no real problems. As soon as I optimized, the errors increased to over 100 pixels. I loaded the successfully aligned photos - before any optimization - into build 1647 and everything behaves fine. I can optimize and refine the project and refine the two unfixed cameras.
I saved and opened the project in build 1667 again and the errors once again jumped from ~0.25 to over 100 on both the fixed cameras. The errors remain at ~0.25 pixels for the other two cameras.
If I go to the camera calibration and uncheck fixed for those two cameras, the errors immediately drop to less than 0.25 pixels for all cameras. If I go back in and check fixed again, the errors jump back to over 100
Tom
-
Hello Diego,
Please try build 1667, now Fixed calibration mode should be working properly.
Hello Alexey,
Now it works perfect, thank you very much.
Best regards,
Diego
-
hello
the pre-release works without any major problems so far, thank you
a small inconsistency when exporting textures was found by me
please find link to google doc zip with all the files so its more clear what it is about
https://docs.google.com/file/d/0B3h8nfSJ97z1MV9TOS1VQ2hSaUk/edit?usp=sharing (https://docs.google.com/file/d/0B3h8nfSJ97z1MV9TOS1VQ2hSaUk/edit?usp=sharing)
photoscan_window_test_no_17.jpg (photoscan 0.9.0)
photoscan_window_test_no_50.jpg (photoscan 0.9.1)
input_test_no_17_1_vs_test_no_50_1.jpg (comparison of input files)
test_no_17.jpg (exported textures on photoscan 0.9.0)
test_no_50.jpg (exported textures on photoscan 0.9.1)
light wise input pictures are pretty the same, no change in camera settings or ambient
window in photoscan 0.9.0 reports blended textures a bit brither
window in photoscan 0.9.1 reports blended textures a lot brither
photoscan 0.9.0 exported textures are similar to the input
but photoscan 0.9.1 exported textures are a lot darker
it could be me missing some option
regards
marcin
-
Hello Marcin,
Thank you for reporting.
Are you processing HDR images or normal ones?
Are you using Standard or HDR Color depth setting in Build Texture dialog?
-
hello Dmitry
yes, the textures were built using HDR indeed
regards
marcin
-
Hello Marcin,
Conversion between standard and HDR data range can result in undesired brightness changes, as it introduces brightness normalization.
In general we recommend to generate textures in HDR mode only if you are processing HDR images, and generate textures with standard color depth if you are processing non HDR images.
Please let us know if you have any specific reasons to use HDR option in the Build Texture dialog.
-
hello Dmitry
there is no reason other than testing new software and all its options
all explained, probably my lack of knowledge here
regards
Marcin
-
Hello Alexey,
I just tried the latest build 1676 with both fixed and not fixed camera calibrations and it is the same as 1667. it does not work properly. I can hold all cameras fixed or all cameras unfixed and have predictable errors. Mixing fixed and unfixed cameras does not work.
Tom
-
I notice that it takes longer for GCPs to show up after I create them in 0.9.1 (build 1676) than in 0.8.5. Didn't use 0.9.0, so I don't know whether the difference was also in that version.
Seems to be taking an average of about 37 seconds for me in a project with four chunks, each with about 250 cameras (12MP images). Model is 50 000 000 faces, system is dual Xeon X5647 2.93 GHz w. 48GB RAM and dual NVidia GTX 560 Ti cards.
with about a hundred GCPs entered that means I spend an hour just waiting for the coordinates to pop up after I click "Create".
(By the way I love that the GCPs stay in numeric order after #9 now when sorted by name/label. But I notice that in the right-click "place point" dialog they don't)
I also love the progress bar when saving a project. Little touch, but much appreciated.
Andy
-
Looks like a bug in the Ground Control\Import dialogue. The row import is incremented by one from what I select. I am guessing that row 1 of the text file = row 0 in the dialogue window.
Anyway, if I export points from one photoscan project and import them to another, the default "start import" row is row 3 (unless I selected that, I forget), with Row 1 being Coordinate System and Row 2 being column headers. But if I select Row 3 to start the import, it skips my first point and starts importing at row 4. I have to select Row 2 in order to import all of my points.
Andy
-
Hi I have issues exporting textures ..when I attempted to do it it says can't export texture.
in version 0.9.1 pre release
I'm running 64bit ubuntu.
Cheers
-
Depth filtering in 0.9.1 (1676) seems to work worse in removing noise and worse in preserving real detail than the default settings in 0.8.5. In my first project done completely in 0.9.1, I tried depth filtering "mild" and "aggressive" and both times had results that were noisier than in 0.8.5 over water areas, and removed real detail (logs on gravel bars) that were preserved in 0.8.5.
Granted, lighting was different, and the flight lines were different, but I have 25 flights I've processed in various conditions in 0.8.5 and subjectively it seems like they are all less noisy over the water than 0.9.1, even in aggressive filter mode.
Is it possible to make the depth filter parameters more customizable than "Mild"/ "moderate"/"aggressive"? I would like to see an option that gives more control - maybe an "Advanced" checkbox that exposes the underlying parameters adjusted by the mild,moderate,aggressive settings so we can see and adjust them more..
Also in the region where the interpolation is happening, is the only way to avoid this to use the "sharp" reconstruction mode? From what I understand, this is not recommended with aerial imagery.
-
Is it possible to make the depth filter parameters more customizable than "Mild"/ "moderate"/"aggressive"? I would like to see an option that gives more control - maybe an "Advanced" checkbox that exposes the underlying parameters adjusted by the mild,moderate,aggressive settings so we can see and adjust them more..
+1
-
Is it possible to make the depth filter parameters more customizable than "Mild"/ "moderate"/"aggressive"? I would like to see an option that gives more control - maybe an "Advanced" checkbox that exposes the underlying parameters adjusted by the mild,moderate,aggressive settings so we can see and adjust them more..
+1
-
I almost posted this in feature requests, then in bug reports, but I think it fits here too.
In 0.9.1 when I try to export a .jpg that's larger than the max. jpg size according to the jpg specification (http://en.wikipedia.org/wiki/JPEG) (65535 x 65535), Photoscan calculates the mosaic size, then continues to render the jpg even if x or y value of mosaic size > 65535.
This wastes a half hour or more generating an ortho that's doomed from the beginning. It seems like Photoscan should abort immediately with an "area too large!" error or something like that once it calculates mesh size, rather than try to generate an "illegal" jpg..
I can't scroll up in the console/log window when it's rendering either, since it updates too fast, so this would be a big help.
-
speed of Pscan
Alexey, this performance im get n smoth and ULTRA, mostly im get aprox 1-1,2 mil how its possible that im get so high results ? 70x18Mpix photos.......
finished depth reconstruction in 45817.6 seconds
Device 1 performance: 198.63 million samples/sec (CPU)
Device 2 performance: 434.339 million samples/sec (GeForce GTX 560 Ti)
Device 3 performance: 871.766 million samples/sec (Tahiti)
Total performance: 1504.74 million samples/sec
-
Hello guys,
I have done some more testing and found some strange behavior of a tool "masking from background".
It might be again my fault or lack of knowledge I am not sure.
I included a zip file with two TIF pictures one with my object and other without it, when I start to import mask from background this is what happens for me
https://docs.google.com/file/d/0B3h8nfSJ97z1ZXlvSnE2alZFMmM/edit?usp=sharing (https://docs.google.com/file/d/0B3h8nfSJ97z1ZXlvSnE2alZFMmM/edit?usp=sharing)
with tolerance set between 1 and 17 the tool does not mask the space between the body and arm
with tolerance set between 18 and 19 the tool does masks the space between the body and arm
with tolerance set between 20 and 76 the tool does not mask the space between the body and arm
with tolerance set higher than 77 the tool again start to mask the space between the body and arm
I might not understand the tolerance setting, anyone could try and see if it happens for him and/or help me understand how to use it properly.
Regards
Marcin
-
I *think* the first part of the "align process, "Detecting points" slowed down (about one 12MP image per second) in build 1684.
I could be wrong, but I thought this used to go much faster than about 1/sec. Images are from two different cameras, with generic preselection and high accuracy setting.
I did just upgrade from two Nvidia GTX 560Ti cards to two AMD HD 7970 cards, and from 48GB RAM to 192GB, but checked openCL and I have two CPU cores disabled. Besides, I think GPU isn't even used in this step? Computer is Dell T7500, dual Xeon X5647 CPUs @ 2.93GHz.
Please someone tell me if it is always this slow and I just am imagining things? I always second-guess myself after a hardware upgrade...
[edit] I confirmed on another machine that this is the speed points are detected. I guess I just never worked with that many images before (or I am getting impatient ) .
Andy
-
Andy OpenCL need to be set with 4 cores disabled - because highly possible you use Hyperthreading, 1 core real + 1 core "virtual" - for every card disable 2 in that settings...
And Aligment work just on CPU for now, but theoreticaly it could be ported to GPU later.... but its not easy task...
-
Should old builds be uninstalled before installing newer builds?
-
Yes. When trying to install a new version and an older version is present on your system, you will be asked to uninstall the older version first.
-
Is the green-blue-gray marker behavior still the same as described in this post (http://www.agisoft.ru/forum/index.php?topic=779.msg3633#msg3633)?
the green flag is used for the projection defined or refined by user, and the blue ones are those placed automatically (using Guided approach for marker placement, for example).
Grey flags are not taken into account for triangulation.
Specifically, is it still true in the 0.9.1 beta that the grey flags are not taken into account for triangulation?
The manual for 0.9.1 says (Page 33), "• If a marker has been placed on at least two aligned images PhotoScan will find the marker projections on the rest of the photos. The calculated marker positions will be indicated with icon on the corresponding aligned photos in Photo View mode."
I am using the GCP import technique described by JMR in the "GCP faster" (http://www.agisoft.ru/forum/index.php?topic=817.msg3747#msg3747) post (which saves me a lot of time, since otherwise I would be entering ~100 GCPs manually every week or two). Generally, after I assign three or so GCPs to two images each, then "update", I can just click and drag the rest of the flags to their proper location. Once I click and drag two flags for the same GCP to the right place, the rest of them are in great locations. But they are still gray until I click on them. I would like to just accept some of those locations as-is (or force the "guided approach" to turn them into blue flags). But if I leave them be and they aren't used for triangulation, that's bad, and if I touch them they move. Is there a way to just "Accept" as-is or force them to behave as in the "guided" approach?
Andy
-
I don't know if this would qualify as a bug or not, but I accidentally left off the minus sign on the one of the horizontal bounds of a jpg orthophoto, which resulted in one of the mosaic axes being 148 million pixels, and the last thing photoscan said before it ate all 192 GB of my memory was:
Mosaic size: 148206540x38093
Estimating mosaic occupancy...
I had to manually terminate the process (I waited ~10 minutes for it to try to stop itself after hitting cancel).
It seems like some kind of reality check on the mosaic size would be a good addition (at least checking to make sure that a jpg mosaic doesn't exceed the maximum legal size of 65535 x 65535 (http://en.wikipedia.org/wiki/JPEG))
-
I am trying to export a dense point cloud in Photoscan 0.9.1 build 1693, and if I leave it alone it does ok, but if I do anything else (even use a browser window) it causes a hard freeze on my computer. I have left computer alone for 4+ hours with it working fine, but as soon as I started using GIS (ArcGIS), I got a hard freeze. Alternatively if I am using chrome when I start to process, it will freeze within a few minutes of starting (last time it froze I had only PS and Chrome with one window filling out a feedback form on AMD's driver website).
Based on the last log entries for each crash, it crashes during depth reconstruction, but it doesn't appear to matter if it is a CPU or GPU operation. These are last log entry from the last three crashes:
[CPU] estimating 4708x3933x2048 disparity using 942x984x8u tiles, offset 0
...
[CPU] estimating 2002x3166x1120 disparity using 1001x1056x8u tiles, offset 0
...
[GPU] estimating 3498x4522x608 disparity using 875x905x8u tiles, offset -608
Project is 272 cameras (12MP), and I am exporting as File/Export Points/ASPRS LAS in local coordinates with point colors, dense point cloud, ultra-high resolution.
I have dual Gigabyte HD7970s NOT in crossfire configuration on a Dell t7500 workstation with 192GB RAM and dual Xeon X5647 CPUs. Each card drives a single monitor on DVI. Power supply is 1100 watt + 450 watt auxiliary. I don't have the cards overclocked, just operating with the default configuration. I duplicated the freeze issue on 13.6 beta 2 AND 13.4 AMD drivers. This issue occurs with both versions.
Symptoms are: Computer goes into a hard freeze. One screen goes blank (not always the same one, but the one that has the active window) and the other freezes. The first time it happened, the screen that didn't go blank maintained mouse movement for a short period, then froze. Once computer freezes, the GPU fans wind down, and the HDD light stops blinking. I force-poweroff with a button push. If I leave the computer alone while processing, it appears to work fine.
I haven't had a problem while generating geometry otherwise.
-
Hello andyroo,
Do you mean that Dense point cloud export and Build Geometry processes have different behavior at depth maps reconstruction step?
-
I'm running version 0.9.1 build 1703. I ran Help>Check For Updates and it indicated I was running the current version, No update required. On the Agisoft download demo web page it list 0.9.1 build 1714. Is this a typo or is it actually a newer build?
-
Hello bisenberger,
There is a newer build with some minor fixes.