wctu / seal Goto Github PK

I'm using Python3.
I encountered ImportError when I imported ERSModule.

ImportError: /home/hongxin/opensource/SEAL/test/ERSModule.so: undefined symbol: Py_InitModule4_64

So, would you please release ERS_Python?

Hi,

Thank you for sharing your code. to use your code I creat a new conda envirenment with python 2.7. Then I installed opencv=3.4.2 and conda install pytorch=0.4.1 cuda90 -c pytorch for cuda 9 and pytorch 0.4.1. However I got RuntimeError: CuDNN error: CUDNN_STATUS_EXECUTION_FAILED Previously I have tried different combination of cuda pytorch and cudnn, still got this error. this is the error and output of python -m torch.utils.collect_env.

Could you please give me a help how I can fix it.

Thanks

It's a great piece of work.
As one of the most important contributions in your paper, will you release the code for the segmentation-aware affinity loss?
Thx ahead.

Hi,

I wanted to run the code with cuda 9.1 on python 2.7, However it was hard to find a pytorch version for python 2.7. Then I have created a new environment with python 3.6. but when running python compile_cpp.py build , I got errors inerspy.cpp related to PyInt_FromLong and Py_InitModule. I have tired to edit the code for python 3, but then I got build\temp.win-amd64-3.6\Release\ERSModule.cp36-win_amd64.lib : fatal error LNK1120: 1 unresolved externals error.

Is it possible for you to provide the corresponding file with python 3.6?

Thanks

Dear wctu,

could you please specify a license for the project? I think github lets selecting one of the predefined license templates and generates a license file in the project automatically (if i'm not mistaken).

Thanks!

Best,
Vasili

Hi,

I use the provided checkpoint on BSD500, and try to reproduce numbers in the paper, but I can only get results:
nC = 200; ASA: 0.959617 BR: 0.733322,
nC = 400; ASA: 0.967188 BR: 0.804939

But from the plot you showed in the paper, it seems you could get >96 ASA when nC = 200 and >81 BR when nC = 400, I'm wondering is this the checkpoint you used in the paper? there seems a slight mismatch with it.

Thanks!

i want to see the testing result based your model, but my pc without gpu..so can this code(testing section) be run in cpu?

Your work is inspiring, is there a plan to publish the training code lately?
Thanks!