Image data ≠ Structured data
Working with image libraries at scale is not the same as working with large sets of structured data.
And this means that statistical tools used to process structured data are inadequate.
How can you find a representative ‘average image’? Or a ‘particularly weird’ one?
But there are visual ways.
How it works
Visually explore large image databases
Gain an intuitive understanding of what images are present in your dataset by browsing 3-dimensional point clouds