Hello,
I am looking into the fast_template_detector package. Could you please
answer several questions I have on the implementation of DOT:
1. I found that DOT from this package is not invariant to small translation
(see section 3.3 of the paper "Dominant orientation templates for real-time
of texture-less objects", S. Hinterstoisser). Is it true? Are there any
plans to add it?
2. It seems that fast_template_detector does not support multi-scale
recognition, i.e. we can detect objects of train size only. Can
fast_template_detector find objects of different size?
3. What are the other differences between fast_template_detector's
implementation of DOT and the original Hinterstoisser's paper or
Hinterstoisser's DOT implementation (
http://campar.in.tum.de/personal/hinterst/index/downloads.html)?
Thanks.
Maria