Memory prolem

81 views
Skip to first unread message

Wen-Han Yu

unread,
Jul 10, 2012, 7:21:53 PM7/10/12
to microbe...@googlegroups.com
Hi
Dose microbeTracker limit the size of the image? I recently run the program to analyze time-lapse images (1392X1040), but often i go the error message about out of memory. I have adjusted matlab java memory to the maximum and the problem was still there. So it seems I need to cut the image into several parts and analyze those individually. Do you have any suggestion? thanks  

Jacobs-Wagner Group MicrobeTracker

unread,
Jun 24, 2015, 12:56:59 PM6/24/15
to microbe...@googlegroups.com, wenhan...@gmail.com
You should be able to run the specified image sizes in microbeTracker without encountering memory problems.  Your memory limitation could be coming from computer hardware such as RAM.

Seán Murray

unread,
Aug 19, 2015, 6:00:50 AM8/19/15
to MicrobeTracker_Jacobs-Wagner Lab
Be sure that none of the meshes are self intersecting.
I have seen a memory warning and also PC freezing due to a self-intersecting mesh segment causing the mex function aip2 (called when integrating the signal) to crash. You can use refine or just delete any problem cells, which you can find by visual inspection (or write a script to detect self-intersections or perhaps just compare the area inside the cell outline to the sum of segment areas using polyarea).

Sean

Jacobs-Wagner Group MicrobeTracker

unread,
Aug 19, 2015, 12:45:34 PM8/19/15
to MicrobeTracker_Jacobs-Wagner Lab
Hello,

Unfortunately, there is nothing that can be done about memory problems at this point.  Please check for updates in the near future regarding this problem.


Thanks,

Jacobs-Wagner Group MicrobeTracker 

Seán Murray

unread,
Aug 20, 2015, 4:44:46 AM8/20/15
to MicrobeTracker_Jacobs-Wagner Lab
I have been working on fixes and improvements to Microbetracker and the associated functions. I still have to implement several of them.
The range from small fixes or small things that would not arise under normal usage (e.g. aip2 only uses the first 4 vertices of the polygon its given) to more involved things that would improve segmentation (e.g. no quality control immediately (same frame) after splitting; the join distance calculation for cell splitting does not take account of overlapping contours, preventing splitting when it should occur; introduce a pre-alignment viewer; add a new pre-segmentation step for microcolonies etc).

Once Oufti is out, I'll see what has been implemented there and, time permitting, contribute code to Sourceforge.

Sean
Reply all
Reply to author
Forward
0 new messages