PCL Developers blog


Here is a detailed outline of my GSoC project milestones:

  • Fill out this page
  • Decide the appropiate benchmark data set for object reconition and pose.
    • CAD models (e.g. Google Warehouse – synthetic data).
    • Data obtained from the Kinect (real data).
  • Implement generic trainer API interface
    • Given training data in a certain format, it should need to load the data and compute the specific descriptor and make all the needed information persistent for the matcher.
    • Documentation.
    • Some descriptors might need specific subclasses to fulfill its specific needs.
  • Implement generic matcher API interface
    • Similar to the generic trainer but for matching purposes.
    • Implement a brute force matcher
    • Implement an approximate kd-tree matcher
    • Implement a RANSAC based matcher interface
  • Benchmarking of existing descriptors.
    • Eventually create a new training set.