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

Finetuning does not 'train' ,aborts #85

Open
ssg5 opened this issue Dec 2, 2021 · 6 comments
Open

Finetuning does not 'train' ,aborts #85

ssg5 opened this issue Dec 2, 2021 · 6 comments

Comments

@ssg5
Copy link

ssg5 commented Dec 2, 2021

I installed the backed server on an Ubuntu 18.04 Virtual machine ( Oracle VM) and i have tried following the instructions in the video for segmentation, finetuning and detection but unfortunately finetuning doesn't work and aborts after showing an error.
I use the binaries provided at https://lmb.informatik.uni-freiburg.de/resources/opensource/unet/caffe_unet_package_18.04_cpu.tar.gz
Finetuning does not even start, aborting with the following

Screenshot from 2021-12-01 18-52-08

The log output :

Screenshot from 2021-12-01 18-52-43

I used the following settings:

Screenshot from 2021-12-01 18-53-23

@ssg5
Copy link
Author

ssg5 commented Dec 2, 2021

I also had a question regarding the tile shape (px) settings. As I am using a caffe framework on CPU I am little confused about the x and y entries for the tile shape (px). I have a CPU of Intel(R) Core(TM) 17-10700CPU @2.90 GHz and a GPU 0- Intel(R) UHD Graphics 630.
What are my upper limits of x and y tile shape(px) I can assign in the finetuning dialogue box?

@ThorstenFalk
Copy link
Collaborator

ThorstenFalk commented Dec 2, 2021

The finetuning problem stems from using the caffe_unet binary which does not support training, for this the regular caffe binary is used. Check your IJPrefs.txt and correct the caffe binary entry to point to caffe, not caffe_unet.

@ThorstenFalk
Copy link
Collaborator

ThorstenFalk commented Dec 2, 2021

Well CPU upper limits are defined by the available amount of RAM, but I would not go too large, it becomes awfully slow.

@ssg5
Copy link
Author

ssg5 commented Dec 3, 2021

Hello Thorsten,
Thanks a lot for your immediate reply, I really appreciate it. I was wondering if you could please explain the exactly what to do to resolve this problem. I am very new to programming and hence i am quite uncertain.
I checked the output of "which caffe" and "which caffe_unet" and attached them below
Screenshot from 2021-12-02 19-32-03

and when i try executing the command "caffe" and "ssh localhost caffe" , this is the output i receive
Screenshot from 2021-12-02 19-32-46

when i was running the sample segmentation, i used to "caffe_unet" and "ssh localhost caffe_unet" commands and got usage messages and while running the sample segmentation for the first time, I THINK i specified 'caffe' in the dialogue box when i got caffe_unet was not found, please specify your caffe_unet binary

image

I really hope you can I help me as I am pretty confused due to my background and experience. Thanks a lot in advance for your help and time.

@ThorstenFalk
Copy link
Collaborator

If you don't need the systemwide installed caffe, I would recommend to deinstall it, then the caffe binary shipped with caffe_unet should be automatically found and used. Otherwise you can edit ~/.imagej/IJPrefs.txt (maybe it is in .config) and just explicitly replace the unet.caffe_binary by /home/kumar/u-net/bin/caffe. If you really don't find it, try find ~ -name "IJPrefs.txt"

@ThorstenFalk
Copy link
Collaborator

Or put /home/kumar/u-net/bin at the beginning of your PATH environment variable, then your system caffe is only found if explicitly called using /usr/bin/caffe.

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

2 participants