Photoscan Error Bad Allocation

[Pendleton115]

I've been trying to build a mesh for an image set of  65 12MP images. When I try to build a mesh I get an error saying Bad Allocation. I've tried raising and lowering the quality, upgrading my RAM and changing machines yet the problem still arises.
PC SPECS:
Intel Core i7 7700k
MSI GTX 980 TI 6G
16GB DDR4
 Advice fixing this problem would be much appreciated. As I'm still new to photoscan and am only just learning the ropes. 

[Alexey Pasumansky]

Hello Pendleton115,

Can you provide the processing log from the Console pane related to the failed operation? Also please provide the screenshot of About PhotoScan dialog - just to make sure that you are using 64-bit version of the application.

[Pholeos8]

I'm having a similar problem.  We have set up a network processing cluster with 16 nodes (see configurations below). 

We are trying to process 260 Cameras.  Each photo is 20 MP from a Phantom 4 Pro.  We are processing the Dense Cloud at "Ultrahigh Quality"

During the "Build Dense Cloud" phase on nodes 1 thru 14, we get the following error messages:

2018-06-20 08:04:23 [146.7.96.111:49778] failed #4 BuildDenseCloud.filterDepthMaps (8/13): bad allocation


When I look at the actual nodes we are getting this series of messages in the command console:
BuildDenseCloud.filterDepthMaps (3/13): point colors = 1
loaded camera partition in 0.719 sec
loaded dense cloud data in 1.243 sec
preloading data... done in 254.701 sec
filtering depth maps... Error: Aborted by user
processing failed in 271.804 sec

The node also displays a Windows dialog that says "Your computer is low on memory. Save your files and close these programs: Agisoft Photoscan"


We are having no problems on Nodes 15 and 16.  The memory usage on those machines is running between 50 and 90 GB while processing.

Is there anyway to get around this "Bad Allocation" issue.  Can the Network Server be configured to send smaller batches to the nodes with less RAM so that they do not run out of memory?

Thanks!


Network Processing Node Configurations:

Processing nodes 1 thru 13:
Intel Core i7-7700 3.60 GHz with 16GB RAM
NVIDIA GeForce GTX 1050
Intel HD Graphics 630

Processing node 14:
Intel Core i7-7700 3.60 GHz with 32GB RAM
NVIDIA GeForce GTX 1050
Intel HD Graphics 630

Processing node 15:
Intel Xeon E5-2640 v. 4 @ 2.40 GHz, 10 cores with 256 GB RAM
NVIDIA GeForce GTX 1080 (8 GB RAM)

Processing node 16:
Intel Xeon E5-2643 v4 @ 3.40 GHz, 6 cores with 128 GB RAM
(2X) NVIDIA GeForce GTX 1080 (8 GB RAM)

[Pholeos8]

Forgot to mention in my previous post:

We are running Photoscan version 1.4.2 build 6205 (64 bit) on all the machines.

[Pholeos8]

Solution:  I increased the Windows 10 Page File Size (i.e., Virtual Memory) for each computer in the Processing Cluster.  In Win10 this setting can be accessed through:

Control Panel --> System and Security -->  System --> Advanced System Settings (left-hand side) --> Performance (Settings) --> Advanced tab --> Virtual Memory Change... --> Uncheck "Automatically manage paging file size for all drives" and set a custom size.  I set the maximum size to be the same size as my installed RAM.  However, that is likely overkill.  You can play around with the settings based on how much free HDD space you have and monitoring the "Committed" memory usage under the "Performance" tab in Task Manager.

After increasing the maximum Page File Size I know longer had any "bad allocation" errors on projects run on individual machines or processing clusters.

[Fkybarless]

Thank you for your prompt reply.

Are you referring only to the dense cloud stage?

I was talking more about the Orthomosiac (final stage) stage.

Your solution looks creative and interesting, and will probably work best on SSD.

[picare]

Solution:  I increased the Windows 10 Page File Size (i.e., Virtual Memory) for each computer in the Processing Cluster.  In Win10 this setting can be accessed through:

Control Panel --> System and Security -->  System --> Advanced System Settings (left-hand side) --> Performance (Settings) --> Advanced tab --> Virtual Memory Change... --> Uncheck "Automatically manage paging file size for all drives" and set a custom size.  I set the maximum size to be the same size as my installed RAM.  However, that is likely overkill.  You can play around with the settings based on how much free HDD space you have and monitoring the "Committed" memory usage under the "Performance" tab in Task Manager.

After increasing the maximum Page File Size I know longer had any "bad allocation" errors on projects run on individual machines or processing clusters.

Hi, I had the same problem and the solution of Pholeos8 worked fine for me.

I have a 70 millions points point cloud, try to mesh it as an "arbitrary 3D" mesh.
System is Dell Precision with i7 7700HQ@2.8Ghz with 32Gb memory and quadro M2200.

Bad allocation while meshing. Fixed the virtual memory to 50 000 Mb (50 Gb) on the SSD drive. That's working fine.
It was fixed at 10 000 before. I don't know why, but it failed.
Thanke Pholeos 8!

[Жандос]

Добрый день!
Проблема такая, при экспорте ортофотоплана в JPEG выходит, Error Bad Allocation.
Пробовал в TIFF все получилось
Но необходимо именно в JPEG
заранее спасибо

[Alexey Pasumansky]

Добрый день, Жандос,

Можете прислать, пожалуйста, лог из панели Консоль, соответствующий экспорту с проблемой и ошибкой, а также сообщить размеры ортофотоплана в проекте (в пикселях)?