Computer Vision for Creative Analysis

I wanted to bridge the gap between performance and creative marketing. As a marketing analyst, I always found it hard to prove quantitatively the impact of art direction or a recurring visual theme , it relied heavily on creative teams being good at taxonomy and able to translate their artistic concepts into measurable parameters.

So I built a computer vision model around exactly the parameters I wanted to see.

To most people, an ad is just a girl in nice clothes staring down the lens. But for someone like me, who lives in binary classification, regression, and correlation — this model gives creative concepts the commercial credit they've always deserved.