search

twke18 / Adaptive_Affinity_Fields

Adaptive Affinity Fields for Semantic Segmentation
https://jyhjinghwang.github.io/projects/aaf.html
Other
262 stars 62 forks source link

How to calculate boundary recall? #8

Closed XiaLiPKU closed 5 years ago

XiaLiPKU commented 6 years ago

In your paper, the detail for calculating BR is not offered. Also, in your code, I can't find any cue. You only offer a citation to Contour Detection and Hierarchical Image Segmentation. But in this paper, the authors refered a way to output the precision-recall curve, instead of a concrete value. So, I just wonder, how to calculate this

jyhjinghwang commented 6 years ago

  1. Please find the evaluation code on BSDS500 dataset website here: https://www2.eecs.berkeley.edu/Research/Projects/CS/vision/grouping/resources.html We did not include this evaluation in the repo as we did not reimplement it but reused their package.

  2. The usual contour evaluation will produce a precision-recall curve as they calculate precision and recall given 100 different thresholds from 0 to 1. However, in semantic segmentation, we don't have or need the intensity of contours. Therefore, there is only one number for precision or recall.