search

costapt / adversarial_retinal_synthesis

22 stars 10 forks source link

request for quality evaluation code ( Mutual Information (MI) , ISC ) #17

Open AliSaeed86 opened 2 years ago

AliSaeed86 commented 2 years ago

Hi @costapt and @diazandr3s, At the beginning, thank you so much for the amazing work you have done and for making the code accessible to all of us. I have learned a lot from you guys, thanks once again.

Secondly, Please if it is possible to share the code of evaluating the quality of the generated images (i.e. Mutual Information , ISC) as i have made some improvement to the results but i don't know how to implement MI measure nor ISC for evaluation.

Thanks in advance. Any help from other readers is highly appreciated.

Regards.

diazandr3s commented 2 years ago

Hi @AliSaeed86,

Thanks for your interest in our work.

The way we measure the quality of the synthetic images was by first segmenting the vasculature and optic disc and then comparing the statistics between real and synthetic ones.

We've also compared qualitatively the real and synthetic ones by plotting 2D histograms and t-SNE plots. Mode details are in Fig 7 and Fig 8 of our paper: Retinal Image Synthesis and Semi-supervised Learning for Glaucoma Assessment - https://ieeexplore.ieee.org/document/8662628

Hope this helps,