About predicted saliency map of HKU-IS in ICON-R

[xuanli98]

Thanks for your great work. I downloaded the predicted saliency map of ICON-R. However, there are only 4445 predicted saliency maps in HKU-IS, instead of 4447. Would you like to provide the full predicted saliency maps of HKU-IS. Thank you very much.

[mczhuge]

Sorry, we also find this mistake before. It seems that many projects share the same problem. Actually, it does not inflect the performance in HKU-IS.

[xuanli98]

Thanks for your response. I have another question: can I fairly compare the performance of each model as long as I use the same evaluation code to evaluate the saliency prediction maps provided by each work? Is it fair to the performance of each model under different numbers of prediction maps?

[mczhuge]

Q1) can I fairly compare the performance of each model as long as I use the same evaluation code to evaluate the saliency prediction maps provided by each work? A1: Yes, if you want to compare the maps given by each work, you should evaluate them under the same evaluation code.

Q2) Is it fair to the performance of each model under different numbers of prediction maps? Actually, not fair! But missing 2 figures will not inflect the performance much, unless there is a deliberate selection of the worst.

[xuanli98]

Q1) can I fairly compare the performance of each model as long as I use the same evaluation code to evaluate the saliency prediction maps provided by each work? A1: Yes, if you want compare the maps given by each work, you should evaluate them under the same code.

Q2) Is it fair to the performance of each model under different numbers of prediction maps? Actually, not fair! But missing 2 figures will not inflect the performance much, unless there is a deliberate selection of the worst.

Thanks for your response !