Open source implementation of KINFU---kinectfusion

Kinectfusion is a Microsoft research project that uses Kinect to reconstruct 3D surfaces in real time and ultimately for human-computer interaction.

Watch the video first: http://research.microsoft.com/en-us/projects/surfacerecon/

Algorithm in this: http://research.microsoft.com/pubs/155378/ismar2011.pdf

Kinfu is an open source implementation of Kinctfusion, developed by the PCL development team and not currently in release.

Look here: http://pointclouds.org/news/kinectfusion-open-source.html

Can't wait to try the effect, then directly compile the code, run to find my video card does not support ....

Kinfu has a high demand for graphics cards, and now there is no formal minimum configuration to publish, the following is a guess to run the configuration:

Nvidia's Fermi architecture graphics, support Cuda, and CC (Compute Capability) above 2.0; Video memory at least 750M;

The students with a strong graphics card can follow the steps to compile, see if they can run (please send the configuration to the comments, thank you):

The detailed procedure under Windows is here:

http://pointclouds.org/documentation/tutorials/compiling_pcl_windows.php

Http://razorvision.tumblr.com/post/14221437255/compiling-the-open-source-kinect-fusion-pcl-point

The following additional brief notes:

My system: Windows 7, VS2010 Pro

1. Code for SVN Repository url:http://svn.pointclouds.org/pcl/trunk

I'm using TORTOISESVN, but it's useless. Google

2. The code version is Rev 5477. Code update is faster, some are not compiled, this version can be.

3. You need to change several settings after running Cmake-gui:

A. CMAKE-cmake_install_prefix//default is in C:\Program FILES\PCL, can be changed to C:\PCL,

Avoid the issue of write permissions.

B. BUILD--BUILD_GPU//hook this up

C. ungrouped Entries--and use_project_folders//If you are using the Pro version of Visual Studio,

This can be checked for easy access to the project.

4. Please install the code before compiling it:

A. CUDA toolkit:http://developer.nvidia.com/cuda-downloads//No restart prompt after installation, manual restart required

B. Update the graphics driver to the latest

5. Compiling "All BUILD" takes a long time (I've spent more than 20 minutes), and there's a couple of Marco issues in the middle.

(This is VS2010 's problem, not resolved), click Cancel to continue compiling. You will be prompted to reload the project at the end of the compilation

Because several new projects have been created. After loading and then compiling the "all BUILD", the new project can be compiled to.

6. Open a command-line window to run your compiled Kinfu so you can see the printing information. My path is

C:\PCL\kinfu_app_debug.exe, the debug version is compiled.

7. If everything is OK, you can see the effect here.

Open source implementation of KINFU---kinectfusion