High-level verification of multi-object segmentation

2016

Closed access Metadata (CSL-JSON)
DOI 10.1109/dt.2016.7557169
Data confidence 70 medium
How this score is calculated

It is built from checkable properties of the record itself: whether the article has a DOI, how many independent sources confirmed its data, whether the required fields are filled in, and how recent the latest confirmation is. Disagreement between sources and a missing identifier lower it. A value of 80 or above means the record was published without further checking; 60 to 79 means it is published at reduced confidence, with a manual check advisable. A record verified by an editor by hand is never pushed below 90 by a recalculation. This scores our data about the article, not the article itself.

Abstract

In this paper we present a relational analysis and verification of multi-labeled images as a result fo semantic segmentation. In semantic segmentation the result is a set of labeled regions but rarely this result is subject to verification that would allow to determine induce the reliability of this segmentation. We show that using shape measure, co-occurrence statistics and specially crafted weighted function we can estimate the correctness of a semantic segmentation up to 92%.