The subset-matched Jaccard index for evaluation of Segmentation for Plant Images

TL;DR

This paper introduces the subset-matched Jaccard index, a new evaluation metric for leaf-level segmentation in plant images that ensures one-to-one region correspondence between predicted and ground truth segments.

Contribution

The paper proposes a novel region-level segmentation evaluation measure that enforces one-to-one region matching, improving accuracy assessment for plant image segmentation tasks.

Findings

The subset-matched Jaccard index effectively evaluates segmentation accuracy.

It enforces one-to-one region correspondence.

It is applicable to leaf-level segmentation in plant images.

Abstract

We describe a new measure for the evaluation of region level segmentation of objects, as applied to evaluating the accuracy of leaf-level segmentation of plant images. The proposed approach enforces the rule that a region (e.g. a leaf) in either the image being evaluated or the ground truth image evaluated against can be mapped to no more than one region in the other image. We call this measure the subset-matched Jaccard index.