regarding label source names in data set imagedatastore

i am working on image classification problem .I am using dataset in which there are some parent folder.Each parent folder have some sub folders.And these subfolders contain the image data.when i use imagedatastore to load ,it takes subfolder as class

I want to take parent folder as class(all the folder in that parent folder sould be a class named as parent folder name).

how to do this .

Answer 1

[Johanna Pingel]

ImageDatastore works really well if you have a parent class and 1 sub folder per class.

imageDatastore('parentfolder/','IncludeSubFolders',true','LabelSource','foldernames');

If not, you may have to do a little more processing. I'm sure there's a ton of ways to accomplish this: I would suggest getting all of your images and labels for each class into a datastore

class1 = imageDatastore({'cat1/random_folder/','cat1/'});
class1.Labels = repelem("cat1",numImages);

Do this for each class, then combine them into one super datastore

imdsFinal = imageDatastore([class1.Files; class2.Files]);
imdsFinal.Labels = [class1.Labels; class2.Labels]);

This should work and be able to customize to what you're trying to do!

[ghada ebrahim]

please i want to know what 'cat1/random_folder/' refer to?

[Johanna Pingel]

cat1 would stand for category 1, as in: a folder with images of category1 in it. This could be any name of your folder for this problem, since this is a hypothetical answer.

random_folder refers to any subfolders which also contain images of category 1.

I believe this was the question posed on how to get all images from folders and subfolders for a category.

[ghada ebrahim]

i get it, thank you so much

[Nilu R Salim]

Hi,

I tried this out. But it's not working.. it repeatedly throws an error. Let me elaborate the issue that I face. I have a main folder with 80 subfolders. Each of the subfolder has 6 more folders in it. I want the labels to be "80". How is it possible?

[Nilu R Salim]

I would really appreciate a quick reply Thanks in advance.

[Nilu R Salim]

I would like to have it something similar to "cd .."

[OJ27]

The code worked for me except for this part

imdsFinal.Labels = [class1.Labels; class2.Labels]);

That line should be replaced with

imdsFinal.Labels = [categorical(class1.Labels); categorical(class2.Labels)]);

Otherwise, you would have a cell array containing strings rather than categorical. After you merge both imageDatastore, then you can call

categories(imdsFinal.Labels) 

to check the labels in the dataset.

[AYUSH ANAND SAGAR]

i want to display the found image's label as the output

how can i do that?

[aref]

labels= unique(trainingData.Labels);

Answer 2

[Neslihan]

Dataset = imageDatastore('Dataset', 'IncludeSubfolders', true, 'LabelSource', 'foldernames');