How to join or merge two Image Data Stores?

18 次查看(过去 30 天)
If I have two image datastores and I wish to concatenate them. How can I do that?
Thanks
  1 个评论
Sherin Aly
Sherin Aly 2018-1-13
编辑:Sherin Aly 2018-1-13
It was a bit tricky but I was able to merge 2 datastores as follows:
imds1 = imageDatastore(location of DS 1,'IncludeSubfolders',true,'LabelSource','foldernames');
imds2 = imageDatastore(imds2 location,'IncludeSubfolders',true,'LabelSource','foldernames');
% create a dummy dataset with the same size of the combined datastores (if ds1 has 100 images and ds2 has 100 images then dummy has 200 images. I made a folder with some photos with equal number of the 2 datastores)
dummy = imageDatastore('Dummy data store locaation','IncludeSubfolders',true,'LabelSource','foldernames');
% note: below this line you can use any datastore variable you want to merge including the ones you get after splitting training and testing
dummy.Files=[imds1.Files;imds2.Files];
dummy.Labels=[imds1.Labels;imds2.Labels];
% Now dummy is the new combined datastore and you can use all datastore functions

请先登录,再进行评论。

采纳的回答

Zhiqiang Qiu
Zhiqiang Qiu 2018-1-22
I know it is an ancient problem, but I just to make it a better reference for other people.
imds1 = imageDatastore('imds1 location', ...);
imds2 = imageDatastore('imds2 location', ...);
imds_combine = imageDatastore({'imds1 location', 'imds2 location'}, ...);
Just combine the address of these two locations. The following parameters are all the same.
  1 个评论
habtamu miheretie
that is cool answer! i have very similar problem. when i create the data store the files are not in the same order as in the folder. how can i make that possible? i need it because the label i use is coming from another .csv file. so i need to match the order to process the images easily. can anyone help me on this please?

请先登录,再进行评论。

更多回答(5 个)

Hana
Hana 2018-1-18
I met the same problem and saw your post. Finally I find out a way:
imds = imageDatastore(cat(1,imds1.Files,imds2.Files))
imds.Labels = cat(1,imds1.Labels,imds2.Labels)
It works for me.
  5 个评论

请先登录,再进行评论。


berkan
berkan 2017-5-22
allImages.Files=[test1.Files;test2.Files]; allImages.Labels=[test1.Labels;test2.Labels];
  1 个评论
Sascha D. Krauß
Sascha D. Krauß 2017-7-3
Unfortunately, this produces a struct instead of a new datastore - and I could not find a way to convert it back. This means, that you cannot use the datastore specific functions like "countEachLabel" etc. Does anyone know another solution?

请先登录,再进行评论。


KSSV
KSSV 2017-2-27
A = imread('ngc6543a.jpg');
B = imread('peppers.png');
%%resize the images so that they have same dimensions
A_resize = imresize(A,[size(B,1) size(B,2)]) ;
%%join them
AB = [A_resize B] ;
imshow(AB) ;

Aasim Khurshid
Aasim Khurshid 2020-12-30
编辑:Matt J 2022-6-28
There is a better way to do this,
imds = imageDatastore({fullfile(imgFolder1Adress),fullfile(imgFolder2Adress),fullfile(imgFolder3Adress)},...
'IncludeSubfolders',true,'FileExtensions','.png','LabelSource','foldernames');

Matt J
Matt J 2022-6-28
编辑:Matt J 2022-6-28
Perhaps by wrapping them in a TransformedDataStore (R2019a and higher),
imdsnew = transform(imds1,imds2, @(x) x)

类别

Help CenterFile Exchange 中查找有关 Deep Learning for Image Processing 的更多信息

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by