Why Image Recognition is Very Useful for In-Store Execution, But Still Little Used.

Augustin Rudigoz

Augustin Rudigoz

CEO
5 november 2018

And How Mobeye is taking it to the next level

Before/After Products Recognition

I meet hundreds of CPG Brands & Retailers every year. And this is mainly to discuss how to spot and correct poor in-store execution as quickly as possible.

Computer Vision (Image Recognition) technologies have theoretically a key role to play in catching physical information from the shelves and convert them into reliable and actionable data.

In short, you send INPUT DATA to your ALGORITHM and get OUTPUT DATA.

Let’s collect a batch of pictures of a given category, stitch them and just send them through a (pick your favorite buzzword)

- Deep Learning Algorithm
- Artificial Intelligence
- Convolutional Neural Network

Getting such KPIs as planogram, shelf share, OOS rate, product price, etc. at a POS level lets Brands and Retailers understand sales data and correct in-store execution in specific flagged PoS.
Here ends the theory.

Most of Brands & Retailers I meet just do not want to hear about Image Recognition. Are they nuts?

Of course not! They are data driven, and results oriented.

Fact is… they consider Image Recognition solutions as a waste of time and money. Ironic for a cutting-edge technology.
Having said that, some retail computer vision technologies work quite well.

So why?

The answer is quite simple, let me correct the previous workflow.

In short, you send INPUT DATA to your ALGORITHM and get OUTPUT DATA.

By saying “Image Recognition is not working” FMCG Brands & Retailers actually say OUTPUT DATA are not satisfactory.

They assume that if OUTPUT DATA are not what they expected, it’s because the Image Recognition technologies are not working well.

FMCG Brands & Retailers opinion

The truth is that some of these technologies are actually working well, but the INPUT DATA is not correct.

We need both good INPUT and well working ALGORITHMS to get good OUTPUT.

What I’m saying is maybe commonplace, but it is amazing to see how many FMCG Brands & Retailers think this way.

The Problem is in the Service

Most Image Recognition companies sell their technology use and ask customers to take pictures (INPUT DATA) themselves. And here comes the problem!

They ask sales team to take shelves pictures with all the requirements of the Image Recognition technology. Overlap, brightness, angle, distance, sharpness… These requirements are quite tricky and if something goes wrong, all the process down the path will be a mess.

This is exactly why a lot of FMCG Brands and Retailers do not fully appreciate Image Recognition Technology.

Here comes Mobeye with one million people-strong Crowdsourcing Force
We really think that FMCG Sales force and Retailers employees should be Sales and Customers focused.

At Mobeye, we combined our Crowdsourcing Service with cutting-edge Image Recognition Technology to create the perfect One-Stop Service for Brands and Retailers.

Mobeye handles both INPUT DATA and Image Processing to give customers the most accurate and valuable data.

Our Million people-strong userbase lets us cover a dozen of countries in Europe, Asia and USA.

Our users take pictures directly with our Mobile App

With this unfair advantage, we can collect the best in-store pictures in a very painless way for customers.

The last step is to process them with the best Computer Vision technologies, where we can use in-house technology or third-party services we have partnership with, depending on the request.

Combining pictures collection & process is the key to take in-store Image Recognition to the next level!

Follow me to get the next articles about Image Recognition.

If you want to know more about our services, feel free to contact us! contact@mobeye-app.com

More articles

Sign up and receive our experts’ newsletter.

By clicking on "Subscribe", you accept our privacy policy.

2013-2019 Mobeye S.A.S All rights reserved