Skip to main content

Advertisement

Table 3 Performance scores on the BSDS

From: Projection-embedded BYY learning algorithm for Gaussian mixture-based clustering

  BSDS500
  Human Mean shift MN-Cut gPb-owt-ucm MML-Jef VB-DNW pBYY
PRI 0.88 0.8157 0.8066 0.7489 0.7851 0.7866 0.8196
VI 1.17 2.2912 2.5163 1.7539 3.4966 3.5589 2.8140
CR 0.72 0.439 0.393 0.439 0.325 0.325 0.487
  1. The performance of each algorithm is evaluated separately against each of five human-drawn ground-truth segmentations per image, and then their average is obtained as the score on this image. For the covering rate metrics, a larger value indicates a better performance.