Image datasets and ML pipelines
When one enterprise has a small quantity of images to handle its easy to keep track and access to these images. You use some labels in form of a text file, keeps all images organized in folders and use the filesystem to allow access.
Once you start to grow a bit we can use something like DVC to organize the data and sync between personnel, rework data and sync between users.
But the problem is, when you start to collect so many data that your team can't have time to possible manage directly or organize in a form that keeps it easy to work.
For that last step I haven't found a good working alternative rather than implement a database with storage linked to the filesystem and a system over It so we can access view and manage the data and metadata.
(WIP)