Agisoft Metashape
Agisoft Metashape => Bug Reports => Topic started by: Eirik Strøm Pedersen on May 02, 2017, 11:21:35 AM
-
I having problems with the new version of Photoscan. The program crash during the build dense cloud operation. Anyone experience some of the same issues? It seems to be related to CUDA or grapich drivers. My grapich drivers are up to date as they was updated last week by the service department in our company last week, so that cannot be the issue.
Right now I'm useing a modded version of Photoscan delivered by Agisoft that solves the issues with running GPU operations with Open CL. But this soulution are just temporary. Attached is the overview of my grapich cards.
-
Hello Eirik,
We are aware of this issue and are fixing it right now. Hopefully the solution will be included in the next update.
-
For those you has issues during depth maps generation on NVIDIA cards using CUDA, I can suggest to check pre-release version of PhotoScan 1.3.2:
Professional: http://download.agisoft.com/photoscan-pro_1_3_2_x64.msi
Standard: http://download.agisoft.com/photoscan_1_3_2_x64.msi
-
Same error here. Wish I knew there was a 1.3.2 version, or that this was a known bug, before I tried 3 times building the dense cloud. :-\
edit: 1.3.2 also failing here.
Version 1.3.2 also fails at the same step as 1.3.1 for me:
2017-05-10 13:04:48 Depth reconstruction devices performance:
2017-05-10 13:04:48 - 4% done by CPU
2017-05-10 13:04:48 - 95% done by GeForce GTX 1080
2017-05-10 13:04:48 Total time: 419.371 seconds
2017-05-10 13:04:48
2017-05-10 13:04:48 Generating dense point cloud...
2017-05-10 13:04:52 selected 102 cameras in 3.623 sec
2017-05-10 13:04:52 working volume: 28412x9927x3742
2017-05-10 13:04:52 tiles: 7x2x1
2017-05-10 13:04:52 selected 5 cameras in 0 sec
2017-05-10 13:04:52 preloading data... done in 0.361 sec
2017-05-10 13:04:52 filtering depth maps... done in 3.086 sec
2017-05-10 13:04:55 preloading data... done in 0.824 sec
2017-05-10 13:04:56 accumulating data... done in 0.764 sec
2017-05-10 13:04:57 building point cloud... done in 0.008 sec
2017-05-10 13:04:57 selected 39 cameras in 0.001 sec
2017-05-10 13:04:57 preloading data... done in 1.957 sec
2017-05-10 13:04:59 filtering depth maps...
-
Hello paulkemp,
Have you submitted the crash report already?
According to the log the issue is not related to the GPU-processing, so it may be a memory issue.
-
Hello paulkemp,
Have you submitted the crash report already?
According to the log the issue is not related to the GPU-processing, so it may be a memory issue.
No, not submitted a crash report other than the logs themselves.
Was able to complete it by installing Nvidia drivers 382.05 clean, removing geforce experiece and running photoscan 1.3.2.
Completed log:
https://pastebin.com/f0LVQMeA
-
Hello paulkemp,
Is the provided log from 1.3.2 related to the crashed operation? It looks like the process has been completed correctly.
If you can reproduce the issue and crash the application, could you please submit the crash report, as it would contain additional information regarding the problem?
-
I'm also getting the same crash on 1.3.2 whilst filtering depth maps.
My Project is aligned on low, c.3700 cameras/images, dense cloud build on low.
Spec:
2 x Xeon E5-2640 v4
2 x Nvidia GTX 1080 Ti
256 GB RAM
Am running 381 nvidia drivers, am trying upgrading to 382.33
-
Hello BristarUK,
Have you submitted the crash report already?
Depth maps filtering is performed on CPU only, so GPU is not used. However, I can suggest to run Memtest86 utility to be sure, that there are now RAM issues.
-
Have you submitted the crash report already?
Depth maps filtering is performed on CPU only, so GPU is not used. However, I can suggest to run Memtest86 utility to be sure, that there are now RAM issues.
Sorry Alexi, I'm against the clock on this one with the model needed by wednesday. So I just reverted to 1.2.6 and all is working again.
-
Just throwing in my two cents...
I was using a GeForce 560 and never had problems with Photoscan
I then installed a GeForce 1080ti and after that had problems with building the dense cloud. Right when it was filtering the depth maps the program would crash or Windows would get an error and the computer would reboot. I'm going to try reinstalling my old video card to see if that's the problem... Sucks if the 1080ti has compatibility issues with PhotoScan.
-
Hello FrostBite,
Depth maps filtering is not utilizing GPU, so any additional information regarding the crashes would be helpful to find out the source of the issue.
-
Hello paulkemp,
Have you submitted the crash report already?
According to the log the issue is not related to the GPU-processing, so it may be a memory issue.
Hi. I am still having issues with 1.3.2 4139. How do I submit a crash report? Crashreporter.exe seems to indicate that it's run automatically. Can you provide a link to an earlier stable version of Photoscan? 1.2.6?
It would be great with some answers regarding these bugs.
-
Hey Alexy,
I was hoping that you're right. But I changed back to my old Geforce 560 and tried the same project again with the same settings and it worked without a problem.
The problem is related to the newer video cards and/or the drivers that are related to them.
Not sure what to do now... it really sucks, since I purchased the new Geforce 1080ti but can't use it for work.
If anyone has any ideas or solutions to this problem please let me know!!
-
Hello FrostBite,
If the issue is reproducible, could you please submit the crash report and put the link to this thread in the Comments field of the crash reported utility.
Additionally I can suggest to use the clean driver installation and check the RAM.
-
Same problem here......in 1.3.2 (downloaded and updated today) with a job of 445 photo's in mode medium and even in lowest mode for dense point cloud generation. It seams to grind to a halt when it needs to preload the data at 89% of the point cloud generation job. Never seen this before. What is strange is that:
a) The times has run down to 0 shortly after the proces hit 89%........
b) The diskload is around 50% continous and that is a pure read proces....no writes.....
Memory is 32 Gb and the load never rises over 40% even in medium load.
Greets, Artisan
-
Hello Artisan,
Can you provide the related log and specify the configuration of your system: OS version, CPU, GPU and driver version?
-
Hello Artisan,
Can you provide the related log and specify the configuration of your system: OS version, CPU, GPU and driver version?
Windows 10 Pro latest build...
i5820k
MSI X99A mainboard (1.09 BIOS is memory serves)
Videocard MSI HTX980Ti driver version.....382.05
32 Gb of 2166 memory
Attached an image of the final screen......inculding the log.....the real log is lost......but this is (and that is good) repeatable.....it happend to mee three times......in a row. You see the videocard to a hollyday and the processor was running nearly idle....only IO was rampant (I waited for an hour).
Workaround for me, I reverted to 2.6.1......waiting for a solution.
Greets, Artisan.
UPDATE, UPDATE, UPDATE!
I got it working. I reverted to 1.2.6, stuck my computer in a Tardis to speed up teh procedures and started calculating. Great.....but since teh Tardis was reconfiscated by a guy called The Docter (which to me was Valentino Rossi) I had to revert again to 1.3.2. So I reinstalled the whole software. And I did not use the MSI software Afterburner......and everyting worked like a charm without a glitch.....so I took 250 of the best pictures checking overlap and let it rip for a high DPC. No problems and a 6.5 million faces mesh, which in resolution outshines a scan made with a 60K Creaform scanner. Now I must say the Creaform scan was a lot faster but qualitywise it's no match anymore.
Keep up the good work........
Arti
-
Same problem here - a mere 112 cameras, works OK in Low quality, but if I try in Medium or High, it blue screens when it gets to "filtering depth maps...".
I've tried with and without GPU, always get the same result.
Win 10
AMD Phenom X6
GeForce GTX 970
32 Gb RAM
It never gets anywhere near maxing out the memory.
Cheers,
Leigh
-
Hello Leigh,
Have you submitted the crash report? If the problem can be easily reproduce, please submit the crash report and put some comment to it, so that we could find it (like link to this thread).
Also I can recommend to check the RAM using Memtest86, as memory failures are quite often reasons of the crashes during the memory intensive operations.
-
Hello Alexey,
I've run Memtest86, it comes back OK. Unfortunately, as the problem causes a "Blue screen of Death", it crashes out before PhotScan can create a crash report, all I get is the log.txt file which I thought I had attached to my last post.
I've run the programme 4 times with different conditions (with/without Gpu, Medium/High) - always crashes at "filtering depth maps..."
I've had the same problem with a different model with even fewer cameras - 67.
All the best,
Leigh
-
Hello Leigh,
Can you make a photo of the BSoD, if there's any information about the address related to the issue? Usually BSoD is related to the hardware issues.
-
Hello Alexei,
I'll get back to you on that one - with Win 10, BSoD disappears rather quickly, so it requires a bit of planning, and a hell of a lot of patience!
All the best,
Leigh
-
Hi Alexei,
Here's a pdf of the BSod, as requested.
I've checked Windows Update and it's up to date, as are all the drivers. I've just run memtest86 from a bootable CD and that passed OK.
I haven't had this problem on earlier versions - I am trying a set of photos which I have run successfully in the past.
Thanks for all your help,
Leigh
-
I had same problem with crash photoscan (stopped drive). Problems was on the soft for overlocking from Gigabyte. I unistall this soft (Xtreme Engine) and everything will be ok. I have GeForce® GTX 1080 WINDFORCE OC 8G (driver 388.71) , MSI X299 Tomahawk AC.
-
I am running 1.4.3 Standard and getting the crashes since upgrading to Nvidia Geforce 1080 (399 drivers). Processing a chunk of 41 images during the alignment process I get error: Kernal Failed:Unkown error (30) at line 228. Previously was able to process the full 178 images on highest setting for alignment. I've disabled Intel HD Graphics 4600 OpenCL card. It doesn't appear to matter.
Log:
2018-09-01 20:04:11 Matching points...
2018-09-01 20:04:11 Using device: GeForce GTX 1080, 20 compute units, 8192 MB global memory, compute capability 6.1
2018-09-01 20:04:11 driver version: 9020, runtime version: 5050
2018-09-01 20:04:11 max work group size 1024
2018-09-01 20:04:11 max work item sizes [1024, 1024, 64]
2018-09-01 20:05:11 Warning: cudaStreamDestroy failed: unknown error (30)
2018-09-01 20:05:11 Finished processing in 261.615 sec (exit code 0)
2018-09-01 20:05:11 Error: Kernel failed: unknown error (30) at line 228
-
Hi there,
I'm having the same issue here with Photoscan Standard 1.4.4.
I'm actually trying some tests to see if it's only in High settings that this bugs happens.
I'm trying right now a medium dense cloud.
My GPU is a Nvidia 1070 with the lasts drivers up to date.
Thanks for any tips that would help :)
-
Just a little more precisions.
So actually my drivers where not up to date, but still, there is something weird going on.
I had lots of kernel crashes with the BSoD.
So I deasctivated the CPU option in the GPU menu, and I just gave Photoscan one core in the affinity menu.
Now, it's been stuck on 'Filtering depth maps" for 4 hours.
Instead of decreasing, the timer is getting higher and higher.
Can someone confirm me if this is normal or if there is a problem ?
My CPU is not that new (i7 2666 gzh) but it's not my first build in photoscan...
Thanks !
-
Hello nicokalo,
Can you please enable saving the log to file via General Preferences tab and provide the output of the failed operation to us (restart the application after setting the log redirection to file)?
BSoD are usually related to the hardware/driver issues, so I can suggest also to check the RAM with Memtest86, as the depth filtering stage is CPU-only.
-
I am running 1.4.3 Standard and getting the crashes since upgrading to Nvidia Geforce 1080 (399 drivers). Processing a chunk of 41 images during the alignment process I get error: Kernal Failed:Unkown error (30) at line 228. Previously was able to process the full 178 images on highest setting for alignment. I've disabled Intel HD Graphics 4600 OpenCL card. It doesn't appear to matter.
Log:
2018-09-01 20:04:11 Matching points...
2018-09-01 20:04:11 Using device: GeForce GTX 1080, 20 compute units, 8192 MB global memory, compute capability 6.1
2018-09-01 20:04:11 driver version: 9020, runtime version: 5050
2018-09-01 20:04:11 max work group size 1024
2018-09-01 20:04:11 max work item sizes [1024, 1024, 64]
2018-09-01 20:05:11 Warning: cudaStreamDestroy failed: unknown error (30)
2018-09-01 20:05:11 Finished processing in 261.615 sec (exit code 0)
2018-09-01 20:05:11 Error: Kernel failed: unknown error (30) at line 228
Hello Griffonaus,
Does the depth maps calculation stage behave in a similar way? Can you try to install older GPU drivers, like 382, for example?
-
Добрый день. Пользуемся пробной версией Agisoft Photoscan, для дальнейшей покупки и использовании в нашей организации. На стадии Обработка-Построить плотное облако точек, при обработке на 95% пишет "bad allocation" и построение облако не происходит.
-
Добрый день, Владимир,
Есть ли возможность предоставить полный лог обработки (строки, включающие в себя пути к проектам можно удалить)?
Также сообщите, пожалуйста, в каком качестве строилось плотное облако, сколько кадров и какого разрешения использовалось, и спецификацию компьютера (CPU, GPU, RAM).
Bad Allocation обычно означает нехватку виртуальной памяти.
-
Добрый день, Владимир,
Есть ли возможность предоставить полный лог обработки (строки, включающие в себя пути к проектам можно удалить)?
Также сообщите, пожалуйста, в каком качестве строилось плотное облако, сколько кадров и какого разрешения использовалось, и спецификацию компьютера (CPU, GPU, RAM).
Bad Allocation обычно означает нехватку виртуальной памяти.
Качество - Высокое
Фильтрация глубины карт - Умеренная
Кол-во кадров - 1307 шт.
Разрешение - 6000х4000
https://yadi.sk/d/fZqzzJwcKnXvNA
-
По логу похоже, что использовалось Ультра Высокое качество, что, естественно, повышает потребление памяти.
Для снижения потребления, можно включить все опции Тонкого Разделения задач (Fine-Level Task distribution) на вкладке настроек сети в окне PhotoScan Pro. В этом случае соответствующие стадии будут при обработке подразбиваться на задачи меньшего размера (аналогичным образом происходит сетевая обработка).
Однако, возможно, также имеет смысл понизить качество плотного облака до высокого.
Поскольку проблема возникла на этапе построения карт глубины и не дошла до фильтрации плотного облака, я бы также посоветовал на всякий случай обновить драйвера для графической карты - на сайте NVIDIA уже доступен драйвер 411.81 для Quadro P2000.
-
По логу похоже, что использовалось Ультра Высокое качество, что, естественно, повышает потребление памяти.
Для снижения потребления, можно включить все опции Тонкого Разделения задач (Fine-Level Task distribution) на вкладке настроек сети в окне PhotoScan Pro. В этом случае соответствующие стадии будут при обработке подразбиваться на задачи меньшего размера (аналогичным образом происходит сетевая обработка).
Однако, возможно, также имеет смысл понизить качество плотного облака до высокого.
Поскольку проблема возникла на этапе построения карт глубины и не дошла до фильтрации плотного облака, я бы также посоветовал на всякий случай обновить драйвера для графической карты - на сайте NVIDIA уже доступен драйвер 411.81 для Quadro P2000.
Добрый день. Удалось создать облако точек по тем параметрам, про которые вы говорили. Теперь такая же проблема возникла на этапе Построение Модели. При этом, из 30 ГБ свободного места на диске (С:) осталось лишь 14 ГБ и чем занято остальное пространство не понятно. Драйвера NVIDIA были обновлены. Логов к сожалению нет, т.к. ПК перезагружался после этого.
-
Появился лог, после очередной попытки.
-
Здравствуйте, Владимир,
Возможно, это файл подкачки увеличился в размерах при попытке расширить объём виртуальной памяти?
Что касается построения модели, что Вы строите ещё в режиме Arbitrary - это очень ресурсоёмкий процесс при использовании плотного облака с большим количеством точек на входе. Возможно, для текущих зада будет достаточно использовать режим Height Field, который строит 2.5D поверхность, либо построить тайловую модель, если интересует высокодетализированная полигональная модель для сложного типа поверхностей.