Every day, millions of photos worldwide are uploaded to social media, so three researchers from Cornell University, Kevin Matzen, Kavita Bala, and Noah Snavely imagined how phenomenal it would be to be able to track global clothing patterns and derive insights, simply by analyzing millions of Instagram photos.
Analyzing millions of photos
Their thinking was that the millions of photos uploaded each day to social media could provide insights into the cultural, social, and economic trends around the globe.
With powerful machine intelligence, they have mined this tremendous database for insights into our society.
Focused on answering one question
The one aspect that the researchers were most concerned about were "how clothing styles vary around the world". Interestingly, this cultural phenomenon has been impossible to study on this scale.
As an example, they wanted to look at the varying use of scarves around the world. With Instagram, the researchers could be specific with the five days of a specific time and within five kilometers of a specific place.
"The team then identified 44 cities to study and downloaded a total of 100 million images from these locations in five-day windows between June 2013 and June 2016."
Step One: harnessing the power of machine learning
Using powerful machine intelligence and an algorithm they developed, they could see how styles of scarves changed over time. There were quite fortunate that the field of machine intelligence is developing very rapidly right now. Using a standard face recognition program to filter out non-face and visible torso pictures, they were left with a set of 15 million photos of people, tagged with their locations and dates.
Step Two: training the machine to recognize clothing
Their next step was training their machine-learning algorithm to recognize various types of clothing and accessories. They taught the algorithm to recognize whether people were wearing a jacket, a scarf, a necktie, glasses, a hat, or something else. The algorithm could also recognize colors, neckline styles, and sleeve length; clothing categories such as T-shirt, dress, or tank top; and clothing patterns, like solids, stripes, plaids, and more.
Step Three: searching for images with similar themes
Finally, using another algorithm, the researchers had the machine search for clusters of images with similar visual themes among the 15 million photos. Their very last task was to track how these images varied over time and across geography.
A whole new way of conducting research
As the researchers concluded: "The combination of big data, machine learning, computer vision, and automated analysis algorithms would make for a very powerful analysis tool more broadly in visual discovery of fashion and many other areas."
Our expectation is that researchers pursuing insights into visual aspects of our civilization will find this new technology very helpful.
© Copyright 1998-2017 by The Herman Group of Companies, Inc., all rights reserved. From 'The Herman Trend Alert,' by Joyce Gioia, Strategic Business Futurist. (800) 227-3566 or www.hermangroup.com
The Herman Trend Alert is a trademark of The Herman Group of Companies, Inc. Reprinted with permission.