Detection skills that returns an image count, properties for cost estimation

[HeidiSteen]

Costing a work load is hard in cognitive search because customers don't know what's inside PDFs or blobs. Since image analysis is so expensive, it would be great to run an initial skillset that returns information about the data itself: image count, image type,

For text, I don't know if this is even possible, but something to indicate what level of granularity would work best for splitting, or suggestions of skills that might add value.