Problem about segmentation fault (core dumped)

[zxprog]

Recently I tried to run the maskfusion s according to the author's steps.But I have But I was entangled in a problem which was showed in the picture below.

My cuda and cudnn versions are respectively 9.0 and v7 as i showed in the following picture

However when I Input nvidia-smi ，it showed that I had a cuda diver 10.1 and I didn't know if this would affect.

the following is my tensorflow version

By the way ,my maskrcnn can run successfully !

I checked it online, the segment fault seems to be an error caused by illegal access to memory, but I have been following the steps.The process of make is also very smooth.

I have been stuck in this place for a long time, I hope I can get help from the big guys.

Thank you very much!
@martinruenz @adrianknobloch

[fovyu]

When using mutiple GPUs, remember to "Select GPU which is running SLAM, this has to be the OpenGL GPU.". Try to change the CMakeLists.txt in maskfusion/Core.

[1171257311]

have you solved the problem?