Extended Summary of the Arc Segmentation Contest

Abstract

Abstract. The Arc Segmentation Contest, as the fourth in the series of graphics recognition contests organized by IAPR TC10, was held in association with the GREC'2001 workshop. In this paper we present the extended summary of the contest: the contest rules, performance metrics, test images and their ground truths, and the outcomes.

A new method to detect arcs and segments from curvature profiles

“…Generally methods are split into three categories [20,22]: Arc-fitting methods [19,8,10] are based on vector descriptions. Main drawbacks are their sensitivity to vector distortion or noise leading to local errors.…”

A general framework for the evaluation of symbol recognition methods

“…Several of these works deal with the evaluation of processes involved in document analysis systems, such as thinning [13], page segmentation [2], OCR [28], vectorization [22,26,27] or symbol recognition [1], among others. In fact, the general performance evaluation framework proposed in this paper is based on the work carried out for the contest on symbol recognition organized during GREC'03 [25].…”

Performance Evaluation of Symbol Recognition

“…In the last years, the graphics recognition community has also become aware of the importance of evaluation and several contests concerning different graphics recognition problems have been organized using the framework of the International Workshop on Graphics Recognition (GREC). These contests have focused, up to now, on the raster-to-vector conversion of line drawings, specifically on the detection of dashed-lines [3], the general vectorization problem [4,5] and the detection of arcs [6]. As a result of this effort, several metrics and protocols for the evaluation of line detection algorithms have been developed [7][8][9].…”