Exported project data contains less samples than labelled tasks and 'remove duplicated tasks' does not work with case sensitive filepaths

[massimodt]

Exported project data contains less samples than labelled tasks and 'remove duplicated tasks' does not work with case sensitive filepaths

[massimodt]

In my case step 3 was unintentional, after labelling some images I had to split them into two projects, so I moved all images into two folders and renamed them, I also splitted the json array and renamed the task.data, unfortunately I missed a lower case letter. After importing the labels, all seemed fine until I found out that there were duplicated tasks and I was labelling images twice. In my user experience it would have helped me save some time if the front end didn't display the images having a 'wrong' case filepath.

Thank you, I didn't know about this feature to modify fields, I managed to fix filepaths in the json file directly and after reimporting it it worked fine.

As a newbie I spent some time to figure it out so I thought that if this happens to someone else this issue might be helpful.

[massimodt]

oh nice, I can filter by 'Storage filename', that is helpful too!

[makseq]

Why do you need this step?

3. rename local storage folder containing referenced data with different case (e.g. .\Images-> .\images)

What paths do you have in your original json file? Images or images?

Remove duplicates will work only if your task.data will be absolutely the same with other tasks. It sounds like you need to fix paths in your database, you can try running label studio this way: EXPERIMENTAL_FEATURES=1 label-studio and try to use Add or Modify Field in the data manager actions. There you can type something like replace("image", "Image") and it should rename your paths. Then you will be able to run remove duplicates.

[makseq]

Also you could filter your tasks by path using Data manager filters and remove all images that you don't need.

[makseq]

Is this issue solved now?

[massimodt]

I think it's still an issue, images should be displayed case-sensitively, it shouldn't happen to have duplicated tasks and 'remove duplicated tasks' does not remove them. Do you agree?

Would it be helpful a pull request to display images case-sensitively?

[massimodt]

in my opinion 'duplicated task' means that it points to the same data source, I think this is the case.

[massimodt]

unfortunately I don't have enough time to contribute to this project, I will close this issue. If anyone else finds useful to fix it feel free to reopen. Thanks @makseq!