PPF3DDetector::read & ::write methods missing

[Algomorph]

System information (version)

• OpenCV => 3.2-dev/master
• Operating System / Platform => Ubuntu 16.04
• Compiler => gcc 5.4

Detailed description

This is kind-of a feature request, because it looks like PPF3DDetector::read and ::write were somewhat intended and defined in the ppf_match_3d.hpp header file but never implemented in the ppf_match_3d.cpp file.

Steps to reproduce

1. Build / install OpenCV with opencv_contrib modules.
2. Write any short program trying to use the cv::ppf_match_3d::PPF3DDetector::read or ~::write methods. The above results in an (obvious) linker error of the form undefined reference to `cv::ppf_match_3d::PPF3DDetector::read(cv::FileNode const&)'

P.S. interestingly, t_hash_int.cpp has hashtableWrite and hashtableRead methods that use the stdio FILE descriptor to read from / write to binary format. This code cannot possibly be adapted for use with FileStorage unless some kind of base64 conversion is used, but if the latter is done, maybe the new WRITE_BASE64 mode can be used on the FileStorage.

[chenshida]

I also meet this problem, has it solve now?

[Algomorph]

@chenshida , seems like it has not been solved (see https://github.com/opencv/opencv_contrib/pull/1094#issuecomment-291527015)

[jibweb]

Are there any plans to implement trained model saving and loading for the PPF matching module? That would also be accessible from the python ?