Google Vision API

We are an advertising agency that wants to test the effectiveness of imagery (logo, colour, Call to Action, Text etc) on the ad.

We collect all ads and their performance metrics from our ad servers like CPC - Cost Per Click, CTR - Click through rate, View-ability, Attention. We want to uploaded these images (ads) into Google Cloud Vision API requesting Label, Text, Landmark, Logo Detection, Face Detection and Object Detection, and stored the results
for analysis.

As the ads are animated HTML5 banners, we want to create a python script that is able to analyses dynamic
animations based on the frame to frame changes and store in Java client libraries.

We want to estimate CTR by using the data obtained by three sources.

1 - Data collected regarding clicks, view-ability and impressions from our ad-server
2 - Information about the presence of objects, labels, faces, logo’s and their characteristics
(position, size and scores) were obtained by Google’s Vision API.
3. Using a K-Cluster algo to calculated shape, sizes, colors, saturation and luminance of the ad.

Outcome:

As we work across various vertical (Retail, CPG, Finance etc), we want to see what is the impact of the following on CTR (click through rate)

- THE IMPACT OF THE USE OF TEXT, CTA and Logo position
- THE IMPACT OF DOMINANT AD COLOR
- THE IMPACT OF Facial Expression
- THE IMPACT OF AD’S MOVING AND THEIR Size