Resizing vs. cropping: current thoughts?

[Nicholas Dufour]

Hi all-

I suspect this might be an inappropriate venue for this, but I was hoping for people to either share their thoughts or point me to current work on resizing versus cropping caffe input data. It is of particular interest to me because I am interested in 'holistic', continuous properties of images (i.e., measurements of memorability and aesthetics) rather than classification and it seems intuitively true that these properties could potentially be nonlinearly affected by cropping and resizing.