Assessment link questions

[DING606]

Thank you for your valuable work!The .mat file is required to obtain the RI, VI and SI indicators. How to get the .mat file?

label_all = load(fullfile(label_path, [ ID, '-planes.mat']));

eval_planes.m code： clc clear close all

pred_path = '../tst/plane_sgmts'; label_path = 'labels/SYN_GT_sgmt';

D = dir(fullfile(pred_path , '*.png')); ri_all = zeros([length(D),1]); gce_all = zeros([length(D),1]); sc_all = zeros([length(D),1]); vi_all = zeros([length(D),1]);

for res_idx = 1:length(D) imgname = D(res_idx).name; pred_res = imread(fullfile(pred_path, imgname));

ID = imgname(1:end-4); 
label_all = load(fullfile(label_path, [ ID, '-planes.mat']));
label = uint8(label_all.assignment);

% based on the experiment it seems upsample the pred to label size will
% lead to a better result
pred_res_upsample = imresize(pred_res,[380 640],'nearest');
[ri_all(res_idx), gce_all(res_idx), vi_all(res_idx)] = ...
compare_segmentations(label, pred_res_upsample);

if min(min(pred_res)) < 1
    pred_res_upsample = pred_res_upsample - min(min(pred_res_upsample)) + 1;
end
if min(min(label)) < 1
    label = label - min(min(label)) + 1;
end

sc_all(res_idx) = compute_sc(uint8(pred_res_upsample), label);  

end

avg_ri = mean(ri_all); % the higher the better avg_vi = mean(vi_all); % the lower the better avg_sc = mean(sc_all); % the higher the better

std_ri = std(ri_all); std_vi = std(vi_all); std_sc = std(sc_all);

sprintf('ri: %.3f, voi: %.3f, sc: %.3f',avg_ri, avg_vi, avg_sc)