Issue with exporting segmentation masks in COCO format

[camkisailus]

Actions before raising this issue

• [X] I searched the existing issues and did not find anything similar.
• [X] I read/searched the docs

Steps to Reproduce

1. Create multi tasks using mp4's.
2. Set some jobs as Train subset and others as Test subset.
3. Annotate some frames with segmentation masks
4. Export the entire project dataset

I then use the pycocotools library to visualize the dataset.

I see that the Train set labels match what was annotated:

But, the Test set labels are totally incorrect:

Expected Behavior

I expect to receive a zip file containing all of the frames in the images directory and three annotation files instances_Train.json, instances_Test.jsonm and instances_default.json in the COCO annotation format. I also expect the annotations and masks to match what is shown in CVAT.

Possible Solution

No response

Context

No response

Environment

- Local CVAT instance running version 2.5.2 on a native Linux machine.