A performance evaluation protocol for graphics recognition systems

Abstract

A b s t r a c t . This paper defines a computational protocol for evaluating the performance of raster to vector conversion systems. The graphical entities handled by this protocol are continuous and dashed lines, ares, and circles, and text regions. The protocol allows matches of the type one-to-one, one-to-many, and many-to-one between the ground truth and the recognition results.

Technical Map Interpretation: A Distributed Approach

“…In fact, in the literature, only the low level operators have been the object of a comparison attempt. Indeed, Phillips et al [56] propose a protocol allowing us to evaluate low primitives extraction. Complete interpretation systems are more difficult to compare, each author basing his own evaluation on his application.…”

Extended Summary of the Arc Segmentation Contest

“…The first contest, held at the GREC'95 workshop, focused on dashed line detection [1], [2], [3]. The second contest, held at the GREC'97 workshop, attempted to evaluate complete raster to vector conversion systems [4], [5], [6], [7]. The third contest, held off-line in association with the GREC'99 workshop, also aimed to evaluate complete raster to vector conversion systems.…”

Performance evaluation of graphics recognition algorithms: principles and applications

“…For this, we propose to use a continuous match value [3,4] rather than a binary, threshold-based decision [1,2,5]. The continuous match value serves to indicate the level of matching of a pair of a ground truth and a recognized object.…”

“…Some researches have been reported on performance evaluation of graphics recognition algorithms [1][2][3][4][5]. However, each of them is for some specific class of graphic objects.…”