fu123456 / SHIQ

Project for CVPR21 paper: "A Multi-Task Network for Joint Specular Highlight Detection and Removal".
85 stars 16 forks source link

No evaluation code #5

Open nachifur opened 2 years ago

nachifur commented 2 years ago

The standard evaluation code is of extraordinary significance to the community. Can you provide evaluation codes for the following indicators? Looking forward to your reply. image

fu123456 commented 2 years ago

For SSIM and PSNR computation, we used Matlab built-in psnr and ssim functions (for example psnr(image1, image 2) and ssim(image1, image2). You can also refer to here https://ww2.mathworks.cn/matlabcentral/fileexchange/64151-structure-similarity-ssim-and-psnr?s_tid=FX_rc2_behav.

fu123456 commented 2 years ago

For accuracy and BER computation, please refer to some shadow detection works such as https://github.com/xw-hu/CUHK-Shadow#cuhk-shadow-evaluation

nachifur commented 2 years ago

Thank you very much for your reply. In SHIQ dataset, are the PSNR=34.13 and SSIM=0.86 in your paper calculated as follows?

for i = 1 : img_num
    i

    image1 = imread([testfnlist_gt{i}]);
    image2  = imread([testfnlist{i}]);

    psnr_ = getPSNR(image1,image2);
    ssim_ = getMSSIM(image1,image2);

    psnr_list(i,1) = psnr_;
    ssim_list(i,1) = ssim_;
end

psnr_mean = mean(psnr_list);
ssim_mean = mean(ssim_list);

image

For SSIM and PSNR computation, we used Matlab built-in psnr and ssim functions (for example psnr(image1, image 2) and ssim(image1, image2). You can also refer to here https://ww2.mathworks.cn/matlabcentral/fileexchange/64151-structure-similarity-ssim-and-psnr?s_tid=FX_rc2_behav.