SOD_Evaluation_Metrics

A more complete python version (GPU) of the fast evaluation for salient object detection (with S-measure, Fbw measure, MAE, max/mean/adaptive F-measure, max/mean/adaptive E-measure, PRcurve and F-measure curve)

• Codes are reimplemented from the matlab version which is available from http://dpfan.net/, modified from metricsource1, metricsource2, and curvesource3.

• Evaluating 1000 images (ECSSD, all metrics) costs about 280 seconds using this implementation, while about 800 seconds for the matlab version. The evaluation results are consistent with the matlab version.

• The source files should be orginized as follows:

  --gt/
        --Dataset_1/
            --img1.png
            --img2.png
                ...
        --Dataset_2/
            ...
  --pred_maps/
        --Method_1/
            --Dataset_1/
                --img1.png
                --img1.png
                    ...
            --Dataset_2/
                ...
        --Method_2/
            ...
  

• Evaluate your map by run:

  python main.py --pred_root_dir ./pred_maps/ --gt_root_dir ./gt/ --save_dir ./score/

• The format of the result file is shown as this.

  #[Dataset_Name] [Method_Name]# [value mae], [value max-fmeasure], [value mean-fmeasure], [value9 adp-fmeasure], [value max-Emeasure], [value mean-Emeasure], [value adp-Emeasure], [value S-measure_alpha05], [value Fbw-measure].

• Draw the PR curve and F-measure Curve by run:

  python draw_curve.py

  Then, the image file will be saved to './score/'. Two virtual curves are shown as follows:

PR curve: Dataset_2	F-measure curve: Dataset_2

• The above metrics are related to the following papers:

  @inproceedings{Fmeasure,
      title={Frequency-tuned salient region detection},
      author={Achanta, Radhakrishna and Hemami, Sheila and Estrada, Francisco and S{\"u}sstrunk, Sabine},
      booktitle=CVPR,
      number={CONF},
      pages={1597--1604},
      year={2009}
  }
  
  @inproceedings{MAE,
      title={Saliency filters: Contrast based filtering for salient region detection},
      author={Perazzi, Federico and Kr{\"a}henb{\"u}hl, Philipp and Pritch, Yael and Hornung, Alexander},
      booktitle=CVPR,
      pages={733--740},
      year={2012}
  }
  
  @inproceedings{Smeasure,
      title={Structure-measure: A new way to eval foreground maps},
      author={Fan, Deng-Ping and Cheng, Ming-Ming and Liu, Yun and Li, Tao and Borji, Ali},
      booktitle=ICCV,
      pages={4548--4557},
      year={2017}
  }
  
  @inproceedings{Emeasure,
      title="Enhanced-alignment Measure for Binary Foreground Map Evaluation",
      author="Deng-Ping {Fan} and Cheng {Gong} and Yang {Cao} and Bo {Ren} and Ming-Ming {Cheng} and Ali {Borji}",
      booktitle=IJCAI,
      pages="698--704",
      year={2018}
  }
  
  @inproceedings{wFmeasure,
    title={How to eval foreground maps?},
    author={Margolin, Ran and Zelnik-Manor, Lihi and Tal, Ayellet},
    booktitle=CVPR,
    pages={248--255},
    year={2014}
  }