Importing tracks turns objects into regular Shapes

[bennet-cv]

My actions before raising this issue

• Read/searched the docs
• Searched past issues

I am trying to import annotations which are tracked objects. Importing those annotations converts all tracked objects into regular objects of the same shape. I only found one issue where other people discussed that they experience the same problem after the issue was closed.
I am currently not quite sure if that feature is not intended, I am doing something wrong or there is a bug so any help is appreciated.

Steps to Reproduce (for bugs)

1. Create tracked objects via running cvat instance and save
2. Export annotations
3. Delete annotations and save
4. Import previously downloaded annotations back into the same job.

Expected Behaviour

Imported annotations are tracked objects.

Current Behaviour

Importing annotations is successful but all objects are regular objects.
I tested in in COCO, CVAT, Datumaro and Pascal VOC datatypes. After the export track_ids, bbox etc. is present in the annotations file.

Context

I want to import automatically created COCO annotations into cvat. But running into big datasets (e.g. 200 images with a few annotations per image) requires a lot of effort if for example one annotation is faulty and I want to remove it (from my understanding I would have to remove the annotation on all subsequent images if it is present). Thus I want to use tracked objects allowing me to just disable it at a certain frame to remove it from subsequent images.

Your Environment

• Git hash commit (git log -1): f578e47
• Docker version docker version (e.g. Docker 17.0.05): 24.0.5 build ced0996
• Operating System and version (e.g. Linux, Windows, MacOS): Ubuntu 22.04.2 LTS
• cvat_sdk: 2.5.2
• CVAT Server: 2.4.6

[zhiltsov-max]

Hi, tracks can only be imported from annotations in CVAT for video (any annotation type except masks), MOT (bboxes), MOTS (segmentation masks) formats, currently. In all other formats they will be imported as separate shapes.

[bennet-cv]

Thanks a lot for the fast reply. With CVAT for Video it works perfectly!