use tensorflow_ros_cpp package without needing to rebuild from source?

[deeptf]

Is there a way I can install this package as any other ros package at ros-kinetic file hierarchy such that I do not need to have sub-module of this package and need to rebuild from source each time?
Have you tried this before?

Thanks again for your response and thoughts.

[peci1]

I was also thinking about that. There are two things that prevent this package from being released:

1. the current mechanism requires this package to be actually built in order to work, since there are no build artifacts it'd provide, just CMake variables and a handful of symlinks.
2. the pip- and bazel-based methods are basically big big hacks (mis)using non-public APIs. I'm not sure if ROS would accept such a package to be released...

The first problem has a solution - rewriting this whole package to work as a CMake "function", module or whatever, that can be called by other CMake files in a similar manner you call find_package. Unfortunately, this idea came to my mind late, when most of the work was already done. So if somebody has the time to do the rewrite, I'll be happy to merge it (it'd be best if it allowed both being used as a catkin package and a cmake module). It's however unlikely that I'll have time for this. On the other hand, most of the "hard work" is already done and it's just about using it differently.