The tests were done on a sample of my photographs as described here
- Tested were 68 of ~1800 images I posted daily over a five-year period
- These 68 each got the most views in the month they were posted
- Each image was “scored” by the MIT algorithms
- A plot of the results is shown here
- Note: the Views axis is for normalized views (actual was many times this)
- The Chart is divided into four quadrants separated by the median values (50% above & below) for memorability (green vertical line) and views (red horizontal)
- Images in quads 2 & 4 support the thesis that memorability implies popularity; Quad 1 & 3 images suggest the opposite
- An actual example image from each quadrant is shown for illustration
- The broken red trend line shows that, in general, more memorable leads to more popular – but I wouldn’t take that to the bank.
An interesting exercise that yielded no surprises. If I were still exhibiting and competing, I’d still go with what “sings to me” (and to the likely audience; Instagram & Flickr viewers have standards that are way different from exhibit jurors and competition judges). My motto has always been – Shoot For Yourself!
This is what happens when you give an engineer a camera 😉