My Thoughts on Technology and Jamaica: How Cloudian and Dentsu smart billboard vehicle recognition takes Facebook and Google on the Road

Tuesday, August 9, 2016

How Cloudian and Dentsu smart billboard vehicle recognition takes Facebook and Google on the Road

Does it bother, please, or amuse you when targeted remarketing by Facebook or Google shows ads for their products?

It's bad enough that Google can target you with ads based on your browsing habits.

However, what if an advertisings platform could identify you based on the model and make of the car you drive?

That the idea behind the smart billboard vehicle recognition platform launching in September 2016 as reported in “Smart Billboards Identify Vehicles to Target Ads”, published June 28, 2016 By Bruce Brown, Digitaltrends.

Google Logo

A partnership between Smart data storage company Cloudian and Japanese advertising company Dentsu, their platform uses cameras mounted on billboards in traffic, big data and deep learning. The image recognition software was able to recognize the make, model, and year of the vehicle 94% of the time.

The billboard displayed a targeted ad on the LED billboard for as long as five (5) seconds, mostly related to their vehicles. This targeted approach to adverting sounds a lot like the CEBS (consumer engagement broadcasting system) that plays targeted audio ads in NCB ATM's as explained in my blog article entitled “How Blue Dot Media CEBS Platform uses IoT to deliver 15 second Audio Ads in NCB ATM”.

So how did an algorithm running on a computer connected to cameras mounted on a billboard are so smart?

Cloudian Dentsu ­ the smart billboard vehicle recognition – Selling Used Cars upgraded

The smart billboard vehicle recognition platform uses cameras and LIDAT to take pictures of the vehicles when they come into view.

It's connected via Broadband Internet to a Server running a Deep Learning algorithm via a data storage cloud provided by Cloudian. The algorithm then analyzes these images and compare them to thousands of used car images. The Database has roughly about 4,000 images per vehicle based on make, model and year.

Google Logo

Once it recognizes a vehicle it then adds that image to its database. This is effectively a form of feedback loop that makes the system so accurate and similar to the AI used by Facebook to recognize people's likes and dislikes as explained in my blog article entitled “How Facebook AI Engineers use AutoML to teach FLOW to make Asimo”.

This process has resulted in the system being able to spot some 200 make and model of vehicles accurately moving at high velocity. It then shows the driver an ad related to their vehicles i.e. old vehicle might see an ad for a newer model and Truck drivers might be shown ads for stops up ahead.

Advertising products to drivers – Minority Report Google and Facebook Style ads coming

It can possible be used to make assumptions about the driver to show them ads specifically promoting products that are often used by drivers of certain vehicles e.g. vacation options for minivan and humvee drivers.

Albeit this ad network may not be effective, due to the speed at which the vehicles move, it can potentially be upgraded with License Plate Scanners to allow Japanese advertising company Dentsu to determine who is driving the vehicle. They can then sell this database to advertisers such as Google and Facebook who can then pay them to advertise their products to drivers.

Minority Report style advertising may be coming sooner than we thought. Except instead of Retina scanners, they'll use Deep Learning and image recognition to recognize.


No comments: