Home/ Questions/Q 8221887
In Process

The Archive Base Latest Questions

Editorial Team
  • 0
Asked: 2026-06-07T14:05:29+00:00

I am trying to create my own Haar Cascade xml file for use in

  • 0

I am trying to create my own Haar Cascade xml file for use in object recognition. By following this tutorial I managed to get it working. However, creating the xml-file takes an extreme amount of time, apparently it can run into a few weeks (I cancelled my own attempt after two days- it was nowhere near complete).

I then came across this group where it seems that this process can be reduced from 3 weeks to 30 minutes by using TBB (Thread Building Blocks). This will speed up the process by utilising all the cores in your system.

The catch is that to use TBB you need to build OpenCV from source, something which I have no experience in. I am following this tutorial, but I do not really understand what I am doing.

I will try and describe my troubles:

1.) Almost all the tutorials mention using Visual Studio to build and configure OpenCV. Does this mean that this OpenCV + TBB setup will only work in the Visual Studio IDE, or will I still be able to do my actual Python + OpenCV (+TBB) development in Eclipse?

2.) At the end the above-mentioned tutorial explains that you can create a .exe file to install your new flavor of OpenCV on other systems. If this is so simple, why has no-one made such an exe (OpenCV + TBB) available?

I hope that this makes sense, I am new to building libraries from source, because I usually stick with the install-files for convenience.

Any advice on how to get TBB working with OpenCV and Python will be greatly appreciated, or even just some clarification of the concepts used in the process.

Extra:

I am using Windows 7 64-bit, Python 2.7, Eclipse Indigo. I do have Visual Studio 2010 installed on my computer (apparently I need a C-compiler).

Share

Leave an answer

You must login to add an answer.

Need An Account,

1 Answer

  1. Editorial Team

    Eclipse uses the mingw GCC compiler? In which case you can just do the mingw build and the resulting binaries will work.

    In any case the build procedure is just to run cmake-gui then either make or nmake on the resulting makefile. The reason all the tutorials are for VS is that VS and Windows users are less used to building projects form scratch than gcc/linux users

    I don’t know if the terms of TBB prevent it being distributed – but it does pick up a lot of processor specific optimisations so opencv would probbaly have to build N versions for different CPUs

    • 0