File does not go into right folder

Hello ,

I am using this code which was written by another user for diabetic retinopathy detection. I am trying to split my dataset to either Yes or No folders, which is in a folder that is created by the code called 'Train Dataset Two Classes\'. The issue that I am facing is that it only writes the image to either only YES or NO. I have no idea why it does this.

Here's the code:

% Two-class Data path
two_class_datapath='Train Dataset Two Classes\';
myFolder = 'C:\Users\adamy\Desktop\MATLAB\train_images\';
% Class Names
class_names={'No','Yes'};
mkdir(sprintf('%s%s',two_class_datapath,class_names{1}))
mkdir(sprintf('%s%s',two_class_datapath,class_names{2}))

% Read the Excel Sheet with Labels
[num_data,text_data]=xlsread('train.xlsx');

% Determine the Labels
train_labels=num_data(:,1);

% Merge all labels marked into Mild, Medium, Severe and Proliferative DR 
% into a single category 'Yes' 
train_labels(train_labels~=0)=2;
% Rest of the dataset belongs to 'No' category
train_labels(train_labels==0)=1;

% Filename
filename=text_data(2:end,1);

% Now, write these images 2-folders 'Yes' or 'No' for us to develop a deep
% learning architecture utilizing Deep learning toolbox
% Determine the Files put them in separate folder
for idx=1:length(filename)    
    % You could uncomment if you would like to see live progress
    fprintf('Processing %d among %d files:%s \n',idx,length(filename),filename{idx})
    
    % Read the image
    current_filename=strrep(filename{idx}, char(39), '');
    img=imread(sprintf('%s%s.png',myFolder,current_filename));
    
    % Write the image in the respective folder
    imwrite(img,sprintf('%s%s%s%s.png',two_class_datapath,class_names{train_labels(idx)},'\',current_filename));
    
    clear img;    
end

Answer 1

[Voss]

Looks like the train_labels you are getting from the first column of that xlsx file are all NaN (and NaN is not equal to 0 so they all get set to 2 and go into the YES group):

% Read the Excel Sheet with Labels
[num_data,text_data]=xlsread('train.xlsx');

% Determine the Labels
train_labels=num_data(:,1)

train_labels = 3662×1

NaN NaN NaN NaN NaN NaN NaN NaN NaN NaN

% Merge all labels marked into Mild, Medium, Severe and Proliferative DR 
% into a single category 'Yes' 
train_labels(train_labels~=0)=2;
% Rest of the dataset belongs to 'No' category
train_labels(train_labels==0)=1;

nnz(train_labels == 1)

ans = 0

nnz(train_labels == 2)

ans = 3662

The solution is to get train_labels from the second column instead:

% Determine the Labels
train_labels=num_data(:,2)

train_labels = 3662×1

2 4 1 0 0 4 0 2 2 1

% Merge all labels marked into Mild, Medium, Severe and Proliferative DR 
% into a single category 'Yes' 
train_labels(train_labels~=0)=2;
% Rest of the dataset belongs to 'No' category
train_labels(train_labels==0)=1;

nnz(train_labels == 1)

ans = 1805

nnz(train_labels == 2)

ans = 1857

[adam yousry]

Yup that fixed it thank you very much. I though NAN meant 0 or non existent so i assumed MATLAB would just skip it,

Thank again

[Voss]

You're welcome!

Answer 2

[Image Analyst]

Exactly what does "split" mean to you? Do you have image files in one input folder, myFolder, and you want to copy (with copyfile) or create (with imwrite) into either a "Yes" subfolder or a "No" folder or subfolder?

I don't like making really long complicated filenames the way you did it:

    imwrite(img,sprintf('%s%s%s%s.png',two_class_datapath,class_names{train_labels(idx)},'\',current_filename));

It's easy to make mistakes that way. Plus you didn't use fullfile() which nicely takes care of whether there is a slash between filename components so you don't have to worry about that. I'd do:

% Create input file name.
baseFileName = sprintf('%s.png', current_filename);
inputFullFileName = fullfile(myFolder, baseFileName);
fprintf('Reading in image : "%s".\n', inputFullFileName);

% Read in input image.
img = imread(inputFullFileName);

% Display input image (optional).
image(img);
axis('on', 'image');
impixelinfo;
drawnow;

% Save to output folder.
outputFolder = fullfile(two_class_datapath, class_names{train_labels(idx)});
outputFullFileName = fullfile(outputFolder, baseFileName);
fprintf('Saving image to : "%s".\n', outputFullFileName);
imwrite(img, outputFullFileName);
% Or alternatively you can use copyfile() instead of imwrite().
% copyfile(inputFullFileName, outputFullFileName)

[adam yousry]

Hi, by split i mean I have a folder which contains all the images, using the excel dataset I have I want to place them in their respectful folder. I have tried your method and it still saves it to only YES. Im assuming its something to do with my logic at

train_labels(train_labels~=0)=2;

% Rest of the dataset belongs to 'No' category

train_labels(train_labels==0)=1;

but im not sure why. In the dataset, if an eye has DR(diabetic retinopathy) it has a value of 1-4 depending on the severity. Although I will admit, your method has made my code look cleaner and i thank you for that

[Image Analyst]

You have to look at your original train_labels and see why they are all non-zero. If you still need help, attach your original train_labels (before you to the reassignments you did above) in a .mat file.

save('answers.mat', 'train_labels')

or (probably better) attach 'train.xlsx'

[adam yousry]

Train.xlss is attached to this and so is num_data. Num_data is the last variable which has all the correct variables before it becomes train_labels.Where all the variables become 2.

[num_data,text_data]=xlsread('train.xlsx');

% Determine the Labels

train_labels=num_data(:,1);

Here is also the code which extracts num_data and places it in train_labels