IMPORTANT: This is a modified version of this repository. I modified the code for my thesis. Read the instructions of the original authors first.

• For my thesis I evaluated 4 models(HS, DRFI, ELD, CPD) on 5 datasets(PASCAL-S, ECSSD, HKU-IS, DUTS Test Set, DUT-OMRON).
• The evaluation measures used were Precision-Recall Curve, F-Measure Score and Mean Absolute Error Score.
• I modified the 'eval_PR' and 'eval_MAE' files so that I could compute the measures for all the models on all the datasets.
  Unfortunately, the ELD model saliency maps on DUT-OMRON are missing 2 files and the original evaluation code is assuming that for every ground truth file there's a saliency map file with the same name: executing the 'eval_PR' and 'eval_MAE' codes then produces an error.
  As time was not on my side while I was doing my thesis, I decided to copy pretty much the same code and skip the missing files in these extra files(I'm aware this is an horrible decision).
  These extra files names end with 'DUT_OMRON'.
• The 'draw_*' files were modified so I could create figures with all the models measures on 1 dataset.
• The 'Unused-Files' folder contains some files that I didn't need.
• The 'FIGURES' folder contains the computed measures images used for my thesis.