R3 rebuttal

[jalabort]

• [ ] Read and comment on the paper:
  • Zhang "Automatic Construction of Parts+Geometry Models for Initialising Groupwise Registration", IEEE Trans. Medical Imaging 2012

[yuxiang-zhou]

This paper introduced algorithms for automatically constructing SDMs for arbitrary articulated objects. It extracts parts from reference frame based on Intensity and SIFT feature. Then matching parts to all candidates on training set. So parts + geometry models can be built. But experiments demonstrates results on fitting X-Ray Scanned objects, which are profiled while our methods focuses on "in-the-wild" scenario. This paper focused on automatically building SDMs by discovering and correlating consistent feature points. While as our research focused on constructing dense SDMs from newly proposed curve annotation.