Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Unable to complete content-based fusion #125

Open
ScientistRachel opened this issue Apr 21, 2023 · 2 comments
Open

Unable to complete content-based fusion #125

ScientistRachel opened this issue Apr 21, 2023 · 2 comments

Comments

@ScientistRachel
Copy link

I am fusing multiview images and wanted to compare the results +/- content based fusion but am unable to run the content based fusion due to time and memory constraints. I understand that the content-based fusion should take more resources than without that option checked, but it is currently taking >24 h vs <3 min, so I wanted to make sure I wasn't doing something wrong, and this wasn't a bug.

I am running the fusion on a Windows machine with 32 cores and 512 GB of RAM. I have updated FIJI to the most recent version of BigStitcher. The raw images I am fusing together are about ~12 GB total in size, which get resaved as a ~5.25 GB hdf5 that if successfully fused ends up as a ~2 GB hdf5. Eventually I would like to apply this to a timelapse, but for now I am only analyzing one timepoint with two channels where there are two angles to be fused together.

If I don't use content aware fusion, the BigStitcher GUI estimates that the required total memory will be ~9644 MB. If I run this fusion while also using the FIJI memory monitor, it ends up using ~16000 MB of memory and the Windows Task Manager thinks it uses ~20% of the CPU resources on the computer. It takes ~3 min per fusion group, so ~6 min total for both channels.

If I check the box for content aware fusion, the BigStitcher memory usage estimate is 21772 MB. Once it starts running, it eventually reaches 400 GB of memory in use (maximum given to ImageJ) and ~100% of the CPU resources in use. I left it running like this for 24 h and it never finished the first group/channel. I closed FIJI at that point, and the hdf5 it had created was only ~30 MB in size after over a day of processing.

Besides the option for content aware fusion, these two tests are the same with the same input images+xml and fusion settings (16 bit, blending true, keep the original anisotropy, fusing to a BDV-compatible hdf5). If I can provide additional information to help pinpoint the issue, please let me know. Thank you for any suggestions!

@artgolden
Copy link

Hi!
I have been experiencing similar problems after updating to the newer versions. I know that version Big_Stitcher-0.8.3.jar and below worked fine. I can send a copy of Fiji with old BigSticher if anybody is interested (since only other way to get it is apparently to compile from source).

@matham
Copy link

matham commented Dec 18, 2023

I'm running into the same issue. I wanted to try an older version of big stitcher (they are available here), and tried installing it manually with multiple older fiji versions (from here), but they all had issues loading stuff.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants