Saliency-dataset-evaluation

November 19, 2019 ยท View on GitHub

This repo can compute the ratio of obj.area and obj.contrast on binary saliency dataset. This Toolbox contains two evaluation metrics, including obj(object).area and obj.contrast.

Note:

  • obj.area: the mean ratio of foreground into masks on dataset.
  • obj.contrast: the mean value of color contrast for all RGB image.

Citation

@article{Saliency-dataset-evaluation,
    Author = {Wei Ji},
    Title = {Evaluation Toolbox for Salient Detection Dataset.},
    Journal = {https://github.com/jiwei0921/Saliency-dataset-evaluation/},
    Year = {2019}
}

Requirement

  • MATLAB

Compute obj.area or obj.contrast

  1. Set the path of dataset in (comput_fore_area_ratio) / contrast_value.m
  2. run .m file

Note: there are two methods to compute obj.contrast. (4-near neighbor and 8-near neighbor)

Saliency maps evaluation

This link near all evaluation metrics for salient object detection including E-measure, S-measure, F-measure, MAE scores and PR curves or bar metrics. The link is here.

@article{Saliency-Evaluation-Toolbox,
    Author = {Wei Ji},
    Title = {Evaluation Toolbox for Salient Object Detection.},
    Journal = {https://github.com/jiwei0921/Saliency-Evaluation-Toolbox/},
    Year = {2019}
}