Recommended/intended way to use Azure Blob Containers as cloud storage in CVAT

[lukaboljevic]

Hey all. I just followed the tutorial for setting up an Azure Blob Container as cloud storage in CVAT, and it all works as expected. However, I'm a bit confused as to what's the intended "workflow" for when you have multiple sets (i.e. folders) of images or videos to annotate (which translates to multiple tasks in a CVAT project).

For example, I have a folder where I keep my data for all my projects. Of course, this folder is nested and has very complex structure. However, I only want to annotate some of the images or videos in specific folders inside.

Firstly, I was wondering how I should upload this data to Azure. Should each specific folder be its own Azure Container, or should I upload everything to one container? Then, I am confused by the manifest.jsonl file. Should this be created for literally all images inside the container, or should it be created per specific folder? Where should one place the manifest file as well?

The manifest file bugging me quite a bit. On paper, its purpose is to reduce the number of requests to the container. The only place to upload or add it is when attaching the container to CVAT. Hence, it could make sense to upload one specific folder of images + their own manifest file in one container. A different folder of images goes in a different container. But then, if you have 10 or 20 folders of images, you have 10 or 20 containers, on top of the fact you have to attach every single one. In other words, this doesn't sound scalable or fun to manage.

Hopefully, I got my point across. So, I can easily connect a container to CVAT and understand how to set up a project/task with the data from it, but I'm confused as to how to use, organize and manage the cloud storage when I may have a lot of folders of images to annotate, i.e. when I may have a lot of tasks per project.

Thank you in advance for the answers!