Dear Mina,

Normally image labeler in Matlab works like ROI (Region of Interest) in ENVI or AOI (Area of Interest) in Erdas to select classes for training a classification algorithm. You can create your own custom vector of image labels by using the following tool

https://www.mathworks.com/matlabcentral/answers/61441-how-can-select-a-region-in-image

I am already using this tool built in matlab the problem when you are going to label images using five classes is time consuming to do this manually?

Dear Mina,

I did not understand your question very well you mean by labeling is to create classes to train a segmentation method? If that is the case than you need to create labels interactively from the image. Otherwise, I would suggest using unsupervised segmentation method with higher number of labels and you can then obtain your 5 classes in different ways. Please correct me if misunderstood your problem.

Thanks for your reply, i am doing labeling for semantic segmentation training

I found using matlab image labeler app in matlab quit easy, to export the images outside matlab you can use the following

imagefiles = dir('.imageLabelingSession_SessionData\*.png');

nfiles = length(imagefiles); % Number of files found

for ii = 1 : nfiles

currentfilename = imagefiles(ii).name;

currentimage = imread(['.imageLabelingSession_SessionData\' currentfilename]);

% images{ii} = currentimage;

% img = double(imread(['Test\' currentfilename]));

im = imread(['.imageLabelingSession_SessionData\' currentfilename]);

v = im; % imagesc handle

map = colormap;

minv = min(v(:));

maxv = max(v(:));

ncol = size(map,1);

s = round(1+(ncol-1)*(v-minv)/(maxv-minv));

rgb_image = ind2rgb(s,map);

imwrite(rgb_image,['labels\' 'rgb' currentfilename]);

% labels a new directory to save your images

end

Credit to my supervisor for the help