SGP 2005 Review Summary

Positive

The paper is sufficiently well organized and written
Labeling scheme is novel
The paper seems to present an innovative contribution on modeling objects through tetrahedral meshes
Paper should be accepted
The presentation of the algorithms is good, the use of examples and diagrams very helpful
The paper appears to be technically sound and relatively well-written
I think the work is potentially interesting. The algorithm can be improved in many ways, including its speed and robustness.
Shape cutting uses an iterative method and is an extension to an exisiting approach
I like that this method allows multiple time intersection of a single tetrahedron using recursive subdivision technique
I applaud the attempt to devise a method to adopt multiple time intersection of a single tetrahedron

Better motivate the advantages of the proposed technique (wrt. existing literature)
Experimental results are not convincing
Some references are missing (3rd reviewer), he proposed three additional ones
Robustness is ignored in this paper (see paper by Bajaj)
The authors avoid of showing the results because he assumes that there are artifacts in the model. Some wired models should be shown
Comment on the limitations of the algorithms and compare it to other approaches
Experiments must be expanded
The bit-wise boundary extraction, point duplication removal with hashing should be simplified or removed (muahh)
More performance tests or comparisons with other techniques should be provided
Not sure about the speed improvements using bit operations. Wondering if a unique ID will give the same performance. Formal performance tests should be done.
Timing chart for shape cutting is missing. 250ms for shape cutting is applicable for real-time applications ?

1st reviewer thought that we do surface cutting
2nd reviewer thought that we use a surface mesh described as a tetrahedral mesh, which is wrong
Concept of wedge is vague (was introduced by Hoppe!)
The time complexity of the algorithm is too high. It needs to touch all the elements for any cut, what appears to be wasteful, since it is fairly easy to do better by adding even minor complexity to the algorithm.
Arbitrarily cutting up volumes does not seem very practical for gaining much information.
On the other hand, binary-tree representation and bit-wise indexing efficiently represent the cutting history information. If such information is more effectively used, the contribution will become more significant. (???)
Topology will also be changed by shape cutting --> no !

Shape cutting should be informally described in the introduction
Limitation of the assumption that 2 pieces are generated by the cut
Ganovelli should be explained more detailed and a comparison should be carried out
What is the domain of the simplicial complex being used ?
Binary sequencing is also called location code
Step-by-step description of the bounary extraction algorithm

Document Actions