G Dorkó, C Schmid - 9th International Conference on Computer Vision …, 2003 - hal.science
We introduce a novel method for constructing and selecting scale-invariant object parts.
Scale-invariant local descriptors are first grouped into basic parts. A classifier is then learned …