MONAILabel Active Learning is not creating new iterations in Endoscopy app

[bilalcodehub]

Hi,

I completed a task in CVAT generated by MONAILabel server. There are 25 images in the dataset folder. The job only picks one image based on the start_server settings. I clicked Save in CVAT UI as well as finished the job. The name of the job is prefixed with DONE_ which means it is marked complete as shown below:

Now, I am expecting a second iteration but the MONAILabel log shows the following message:

[2023-02-11 09:19:21,024] [2012] [MainThread] [INFO] (lib.scoring.cvat:76) - Latest Active Learning Task in CVAT is under progress/not-consumed. Skip to create new!

This means that MONAILabel is not informed of the completed task. What I will be doing wrong? I tried checking CVAT UI options but didn't find anything useful. Do I have to upload annotations to let the label server know about this?

[bilalcodehub]

I see that MONAILabel is creating a mask for the completed task inside datasets folder under labels/final path. If I download masks from CVAT it properly labelled as done by tooltracking model, but the one added to the datasets folder by MONAILabel is blank mask containing black pixels only as shown below:

Mask downloaded from 'CVAT' using Export dataset option (though images are different):

[SachidanandAlle]

these are PNG images which label values 1,2,3 kind.. so you will not see them in normal viewers.. you can try to load them in 3D slicer or MITK workbench or ITKSNap kind.. the value of pixels (labels) are between 1-255