Read images from folders in directory - matlab

Question

1) I have an original directory called "Original_file" that contains several number of images. The code below serves to read those images from the directory, converts them to greyscale, then write them into new directory called "Target_File".
Target_File='modified_images';
mkdir(Target_File);
directory='original_images';
dnames = {directory};
cI = cell(1,1);
c{1} = dir(dnames{1});
cI{1} = cell(length(c{1}),1);
for j = 1:length(c{1}),
cI{1}{j} = double(imread([dnames{1} '/' c{1}(j).name]))./255;
cI{1}{j} = rgb2gray(cI{1}{j});
imwrite(cI{1}{j}, fullfile(Target_File, ['image' num2str(j) '.jpg']));
end
2) From the "Target_File": The code below serves to select randomly a specific number of images and put them in a training file.
Train_images='training_file';
mkdir(Train_images);
ImageFiles = dir('Target_File');
totalNumberOfFiles = length(ImageFiles)-1;
scrambledList = randperm(totalNumberOfFiles);
numberIWantToUse = 5; % for example 5
loop_counter = 1;
for index = scrambledList(1 :numberIWantToUse)
baseFileName = ImageFiles(index).name
str = fullfile('Target_File', baseFileName);
image = imread(str);
imwrite( image, fullfile(Train_images, ['image' num2str(index) '.jpg']));
loop_counter = loop_counter + 1;
end
What I want in this question ?
A) If we consider that we have a directory that contains several number of folders (folder1, folder2, ...., foldern). Each of these folders contains several images. So how can I edit my code in 1) in order to apply the same concept and get a new directory "Target_File" that contains the same number of folders, but each folder becomes containing the greyscale images?
Then, from the Target_File created in A) : I want to select (randomly as in 2)) from each folder in Target_File, a specific number of images and put them in training file, and the remaining images in testing file. This procedure is repeated for all folders in the directory.
So if the directory contains 3 folders, each of these folders is split into training and test files. So the first folder is split into train1 and test1, the second directory into train2 and test2, the third directory into train3 and test3, etc. So how to edit my code in 2) ?
Any help will be very appreciated.

Answer 1

You can use the dir command to get a list of sub-directories, and then loop through that list with calls to mkdir to create each one in turn. After that, it is just a matter of matching the file paths so you can save the greyscale image loaded from a source subfolder to its corresponding target folder.
Specifically, D = dir('directory') will return a struct where each element of the structure is an element stored in 'directory'. D(i).isdir will be 1 if D(i).name corresponds to the name of one of your subfolders (note that you will need to ignore D(1:2), as those are the folder navigation tags . and ..). So, get your list of directory contents, and then loop through those calling mkdir if D(i).isdir is 1.
I am not sure I understand the rest of your question, but if you just need a random subsample of the entire image set (regardless of subfolder it is stored in), while you are making your subfolders above you can also make secondary calls of dir to the subfolders to get a list of their contents. Loop through and check whether each element is an image, and if it is save it to an array of image path names. When you have compiled this master list, you can grab a random subset from it.