PCL Developers blog

Ioan Dumitru Dragan

This is my personal page

project:Real-time segmentation and tracking
mentor:Nico Blodow (Radu B. Rusu [Suat Gedikli, Michael Dixon])

About me

I’m a student at West University of Timisoara, Romania.

My hobbies include:

  • Coding
  • Fishing
  • Snowboarding
  • Reading


First, read this description of the current pcl_cuda codebase for an overview.

Here is a brief outline of my GSoC project milestones. Click here for a more detailed roadmap.

  • Fill out this page, discuss road map with mentor(s).
  • Get to know the current code base for pcl_cuda.
  • Documentation: For the remainder of the milestones, while developing, make sure to document all code to allow the mentors to easily follow your train of thought and progress.
  • Together with the mentors, put in place a system to evaluate performance and correctness of the implemented methods.
  • Port a seleciton of the existing parametric models and sample consensus algorithms variations to the new framework.
  • Implement a bridging layer between OpenCV GPU and pcl_cuda where we can convert from a structured pcl_cuda::PointCloudAOS to cv::gpu::GpuMat and vice versa.
  • For true 3D cluster segmentation, implement a graph-based region growing segmentation algorithm (see pcl::EuclideanClusterExtraction).

Recent status updates

Monday, July 04, 2011

Last week after some discusions with Nico, I managed to correctly complie and install OpenCV with CUDA support. This part was messy for me. After that I recompiled the source code of PCL, WITH_CUDA, and compiled the kinect_ransac.cpp in order to get more familiar with the framework. During the weekend I started to implement the MSAC algorithm. I can say that here I have progressed a little. The code is not done yet, but is due tomorrow night. Today I am planing to implement and test the sac model functions needed to be able to compute the MSAC. - this is inspired from pcl/sample_consensous/. The next post, probably tomorrow night will contain some results from the MSAC implementation with CUDA.

Progress report
Monday, June 20, 2011

Due to some minor problems I was not able to work on gsoc last week. Today I downloaded (knowing that Nico had updated the Cuda code last week) and compiled the updates concerning the CUDA code. I started to get a general ideea on what is going on. The plan is that this night I will try to better understand what was done. And tomorrow to test some of the code on my machine. After this I have to get some guidance from Nico concerning what is to be done next.

Progress report
Monday, June 13, 2011

This week I started by reading the document provided along with M-estimator SAmple and Consensus. (MLESAC: A new robust estimator with application to estimating image geometry, by P. H. S. Torr and A. Zisserman.) I also started reading and testing the existing RandomSampleConsensus class. At the moment I am trying to figure out how to implement this MSAC using the existing CUDA code. As I progress I will post more details concerning the implementation.

Wednesday, June 08, 2011

Last week I have been very busy- reason for no posts. I started getting familiar with the THRUST library and with the project. As a quick todo, this days I have to get in contact with Nico and discuss more details about the roadmap. As a plus side of the last week+now: I am participating at a HPC computing ( Toward petaflop numerical simulation on parallel hybrid architectures - INRIA sophia antipolis) summer school where I have the great opportunity of listening to Jack Dongara and other interesting persons. As far as gsoc concerns me, starting from this week I will have enough time to perform all the necessary tasks, as I finished with the work needed for my master thesis and for the internship.

Status update week 1
Monday, May 30, 2011

I finally got PCL compile on my machine. The progress I’ve done up to now:

  • Installed all the dependecies (except openni - still working on that)
  • Changed several times the Ubuntu distribution - so that I have no nvcc problems
  • Installed the cudaSDK, devDrivers and so on
  • At the moment I started checking the existing CUDA code

This week will be a full week for me due to ending of the INRIA stage.