Hope it's ok to suggest this here (thought it more appropriate than the forum)...

I'm an opencv n00b, but have years of C/C++ ...

Tried this week to build ocv for the first time. The goal was to run at least the 'show Lena' code. I've still not got it working ! ... tried on Fedora, Ubuntu, and Mac. All clean, up-to-date platforms. In addition to a simple GCC run, I also tried using the CDT/Eclipse approach on all three platforms. So that was 3 x 2 = 6 versions x 2 approaches = 12 failed attempts. The failures were at all stages for various reasons. I see the main problem as vague or incomplete documentation.

The docs for Mac/Linux need to explain where all the files go at the various stages of the build - I found little to no documentation on this crucial topic at all, but did come across many posts all over the net where people were looking for such locations. It would have saved me hours of headaches to have this at hand, as in a big, easy-to-see link from the home or doc home page:

Call it a preflight check list for getting Lena to show her lovely face. Some of this is 'sort of' explained, but not good enough if something fails. ( Got worn out using 'locate {filename}' )

0. put external libs we need {here} // e.g. libavcodec-dev, etc.
1. If you put git or tz file {some_dir}
2. Then unpack to {there}
3. You will find cmake files {somewhere_here_or_there}
4. After cmake, you will find logs {golly_only_knows}
5. The libs generated by cmake, such as libavcodec-dev, will go {here_lib}


6. Try running some sample code (... And *please* make sure it was built from scratch using this very check list ! )
7. If you get 'references not found' (unknown / undeclared) errors, then look back to {where_they_are} to make sure .h / .hpp are there.
8. If you get linker errors, or unknown references after an Eclipse build, look at {where_ever} to make sure .so files are there.

Happy ending with Lena :))

thanks
- Howard in Florida