search

yan99033 / monodepth-cpp

Tensorflow C++ implementation for single-image depth estimation (Monodepth)
MIT License
105 stars 33 forks source link

Should I install tensorflow in virtualenv method instead of source code method? #2

Closed zxp771 closed 5 years ago

zxp771 commented 5 years ago

Thanks for your sharing. As the title said, I tried to install the tensorflow in virtualenv but it doesn't work when I compile the monodepth. the error like this:

/home/zxp/monodepth-cpp-master/src/monodepth.cpp:30:41: fatal error: tensorflow/cc/ops/array_ops.h: No such file or directory compilation terminated. CMakeFiles/monodepth_static.dir/build.make:62: recipe for target 'CMakeFiles/monodepth_static.dir/src/monodepth.cpp.o' failed make[2]: [CMakeFiles/monodepth_static.dir/src/monodepth.cpp.o] Error 1 CMakeFiles/Makefile2:67: recipe for target 'CMakeFiles/monodepth_static.dir/all' failed make[1]: [CMakeFiles/monodepth_static.dir/all] Error 2 Makefile:127: recipe for target 'all' failed make: *** [all] Error 2

Is this caused by the virtualenv method won't install the bazel?

yan99033 commented 5 years ago

If you install Tensorflow using pip install, it won't work.

You need to build and install Tensorflow from source, check this link for instructions. Furthermore, you can find more information by clicking the links (under Reference) down below the page.

zxp771 commented 5 years ago

@yan99033 Thanks for your reply. I have read your instruction. BTW for my GPU the drive for CUDA 9.0 is too old. So I have to install CUDA in 10.0 and CUDNN in 7.4 I keep the bazel in 0.11.1. Do these work for your code?

yan99033 commented 5 years ago

Hi zxp711,

I don't think CUDA toolkit 10 will work. Tensorflow officially supports CUDA 9.0; you would need to downgrade your nvidia driver and install the compatible CUDA and cnDNN versions.

Check out the Tensorflow link to find out the compatible versions for CUDA, cuDNN, and Bazel tools.

zxp771 commented 5 years ago

@yan99033 Thanks for your suggestion. I'll try it in cuda 9.0.