undefined reference to tensorflow::readbinaryproto error

Question

undefined reference to tensorflow::readbinaryproto error

JustWon opened this issue 6 years ago · 5 comments

Hello, I am trying to compile your newer version of segmap which containing the cnn descriptor.
the most of catkin packages were built but I encountered an error while "catkin build segmapper" as below.

I searched for this issue on Google and found out that it's because the dynamic lib of tensorflow is not installed.
mgharbi/hdrnet_legacy#10

so I re-built tensorflow 1.8.0 with the below command and re-installed.
bazel build --config=opt --define framework_shared_object=true tensorflow:libtensorflow_cc.so

However, the same error happens.
Can you guess what is the problem?

Thank you,

Answer 1 · 2018-09-06T12:11:56.000Z

Seems to be a tensorflow compilation issue. What you could do is try with a newer version of tensorflow, such as 1.10. Only thing you'd need to do after that is retrain the models which are not compatible across versions of tensorflow.

Answer 2 · 2018-09-07T01:43:02.000Z

@smauq Thank you so much!
I successfully built the newer version of Segmap with Tensorflow 1.9.0!

Answer 3 · 2019-03-26T13:18:31.000Z

Hi,I also install the Tensorflow 1.9.0,but when I run 'catkin build tensorflow_ros_cpp --cmake-args -DFORCE_TF_PIP_SEARCH="ON"' it turned a error: Tensorflow was not found
What can I do to solve it ?
Thanks

Answer 4 · 2019-08-07T08:24:23.000Z

Hi, @JustWon and @smauq . After so many times trying, I solved many issues but still got the exact same problems as this one. I use TF 1.10 and compiled it from the source using command below:
bazel build --config=opt --config=cuda --cxxopt="-D_GLIBCXX_USE_CXX11_ABI=0" --define framework_shared_object=true tensorflow:libtensorflow_cc.so //tensorflow/tools/pip_package:build_pip_package

Could you please give me some details about how to solve this problem?

Answer 5 · 2021-10-08T07:55:05.000Z

Hello, I am trying to compile your newer version of segmap which containing the cnn descriptor. the most of catkin packages were built but I encountered an error while "catkin build segmapper" as below.

I searched for this issue on Google and found out that it's because the dynamic lib of tensorflow is not installed. mgharbi/hdrnet_legacy#10

so I re-built tensorflow 1.8.0 with the below command and re-installed. bazel build --config=opt --define framework_shared_object=true tensorflow:libtensorflow_cc.so

However, the same error happens. Can you guess what is the problem?

Thank you,

hi,I encountered the same problem with you,can you tell me the answer?