25

What is it?

If you want to run analytics on images then you need to have the images to analyse. This process is therefore the dedicated collection of useful images for that purpose.

With the increase of social media platforms and smartphones there is now an abundance of images that could potentially be useful for analysis.

Why does it matter?

Image data matters because a picture paints a thousand words and that is still true. There is a huge amount of information that can be gleaned from a photograph or image above and beyond what is contained in the image. For example, photographs have GPS-based meta-data that will tell you when the photograph was taken as well as where the photograph was taken.

How can I use it in practice?

Image data can be very useful in manufacturing and improving customer service. If a part is faulty, for example, a picture of the fault can often help the manufacturing team to understand the problem better. If all faults are always recorded with a photograph of the problem then it’s more likely that consistently faulty parts can be identified and improved.

Insurance companies often now ask for photographs of the incident or damage that is initiating an insurance claim. In addition, when a customer is unhappy with a product or service they can be asked to send a photograph of the fault to fast track the return process.

Crime prevention has been using image capture data for a long time, taking multiple pictures of a crime scene to record as much data as possible.

Once you have the image data you can apply image and video analytics (Chapters 10 and 11) and sentiment analysis (Chapter 9).

How do I get started?

The best place to start when gathering image data is to identify what image data you already have and whether it’s going to be useful. Also, as always be very clear about what questions you want your image data to answer.

Consider how images and photographs could help you to improve your products, services and customer service. This is especially true now as so many people have smartphones and are constantly walking around with a camera in their pocket. Asking for a customer or employee to take a photograph of a fault or incident is no longer an imposition. If a photograph will be useful ask for it.

Also, if you have CCTV footage for security purposes this can be easily converted to image data.

Possible data sources

There are many possible image data sources including:

• photographs;
  
• graphics;
  
• online images;
  
• video footage;
  
• CCTV footage.
  

How difficult or costly is it to collect?

Whether you already have access to image data or not, most people have constant access to a good-quality camera via their smartphone so if you think image data could help you to improve your products or services then ask your customers for photographs.

You could run online competitions or ask via feedback areas on your website so customers can upload photographs of faults or new and novel ways that they use your product. This type of insight could lead to new product developments and improvements.

Practical example

The king of image analytics is probably Facebook! Billions of people use Facebook and happily upload 350 million images every day!¹ To say they have a lot of image data would be a staggering understatement – all provided for free by us. The result is a massive image library which has allowed Facebook researchers to create their DeepFace pattern recognition system capable of achieving near-human face recognition accuracy. The software automatically converts the topography of each face in the image gallery into a unique mathematical code, called a faceprint. This faceprint can then be used to identify an individual from other photographs online, subsequent photos uploaded to social media sites or CCTV footage.

Using the photographs users willingly upload, Facebook is able to create a 3D image of their face and then scan the rest of the internet to find other pictures of those people to see where else they appear and who else they might know. So, for example, Facebook can and are matching your uploaded Facebook images to your corporate website, online dating profile or find you in news articles or blog posts. This way companies are able to triangulate the data they have on you and find out even more. There will come a time in the not-too-distant future when Facebook will be able to monitor your photographs and notice if you have put on weight. That data could then be sold on to weight-loss companies for targeted advertising.

Tips and traps

Encourage your customers and employees to record images and send in photographs of your products and services in use. Build up a database of images that can then be analysed to discover trends. For example, you may discover that while one particular product is sold in a particular store it is being used at a different location, which could indicate a potential new market.

Remember the time when photographs only existed in physical form? While they may be useful for retro marketing or illustrate how your products have changed over the years don’t get too hung up on converting all your old physical images into digital form – unless there is a specific reason for it.

Further reading and references

To find out more about image capture see for example:

• Baughman, A., Gao, J. and Pan, J.-Y. (eds) (2015) Multimedia Data Mining and Analytics: Disruptive Innovation Hardcover, New York: Springer
  
• Shan, C., Porikli, F., Xiang, T. and Gong, S. (eds) (2012) Video Analytics for Business Intelligence, New York: Springer
  
• Daoudi, M., Srivastava, A. and Veltkamp, R. (2013) 3D Face Modeling, Analysis and Recognition, Hoboken, NJ: Wiley
  

_________________

¹IACP Centre for Social Media Fun Facts http://www.iacpsocialmedia.org/Resources/FunFacts.aspx