Image classification pipeline

[dhrogers]

Initial use case is an image classification pipeline. This takes as input a cinema database of images, uses a previously trained machine learning model to classify input images. Classifications are added to the image metadata, and a table is created from that.

A table query makes sense if the resulting table incudes image objects, instead of URIs to resources that must be reloaded.

Questions remaining:

1. how is the resulting table used to filter images in a final image view?
2. are the resulting images re-loaded, or can/should the previously loaded images be used?

[dhrogers]

@JonasLukasczyk, @dbanesh: See diagram above from our discussion today. I'm not clear on how images get passed to the Image view without reloading them. Perhaps that's not important in the first pass, but worth thinking about for large datasets.