RANVEC and the Arc Segmentation Contest: Second Presentation

This paper provides some information regarding the winning system at the GREC’2005 contest on arc segmentation. Important facts are first recalled, then the changes made on the system since its first presentation at GREC’2001 are detailed. The obtained results are briefly commented, and the paper finally provides some clues about possible, future improvements regarding the system. 1 Some historical notes The system mentioned in this paper was first presented during the GREC’2001 workshop, and took part of the arc recognition contest held at the same time in Kingston, Canada. At that time, results from the competition were not really enthousiastic [6]: the average VRI value did not exceed 0.63, many arcs were misdetected, and the presented prototype even crashed on an image. Although the reader may find it suprising, the method has almost not changed since that time; but its implementation definitely has. Indeed, to the exception of what is presented in the next section, the system described in this paper strictly follows what is detailed in [3] and [2]. It has been reimplemented only recently as a 64-bit PowerPC application, and actually runs on any Apple computer fit out with a G5 processor. The main goal of this short paper is therefore to inform the reader that the average results obtained in 2001 are mostly explained by a poor implementation of the method, not by an intrinsic default in the method itself.