insights

4 Colorful Applications for Computer Vision

As marketers, color is one of our most basic—and most important—tools: Studies have shown that up to 90 percent of a person’s initial impression about a product is based on color alone. Yikes, right? The good news: In recent years, computer vision technology has become more sophisticated, enabling machines to “see” the whole spectrum—and allowing us to harness the physiological power of color with the equally impressive might of machine learning. Here are four applications for the rosy future of full-color computer vision:

1. Making a Match
British home improvement retailer B&Q found that customers weren’t fully utilizing its new paint mixing service that can match a paint color to literally any object larger than a fingernail. Enter GumGum’s computer vision, which scanned editorial photos as consumers viewed them online, analyzed millions of pixels, then created customized in-image ads perfectly color matched to a prominent object within in the photos—all in real time. (Watch a video of the ads in action here.) The result? Once potential customers saw this demonstration of how they could have walls, say, the same mauve as Manolo Blahnik’s coat, B&Q boosted its association as both a place to buy paint and as the spot for paint matching services.

 

2. An Eye for Fashion
We already know that computer vision is remaking the fashion industry, and the ability for machines to see in color is leading the charge down that runway. In a 2017 study, color was among the primary attributes Cornell University researchers used for analyzing 15 million shared photos in order to detect global fashion trends. Condé Nast also used color as a key component of its prototype handbag classifier, part of a robust machine-learning program the media company uses to analyze its published content. By scanning about 17,000 fashion images from selfies to professional shots, the computer model was able to correctly identify a handbag’s color and brand about as successfully as a human handbag aficionado—paving the way to a better understanding of the insights imbedded in Condé Nast’s trove of images.

 

3. Not-So-Basic Beauty
Selecting a shade of foundation is a lot like picking out a new pair of jeans: If it’s not just right, you might as well plan to stay in the back of the closet. Sephora has recently tackled this age-old beauty bind with an AI-driven device called Color IQ, which scans customers’ skin tones and assigns a Color IQ number that suggests foundation, concealers and other makeup from the thousands of shades available. This in-store offering joins the mobile tool Sephora Virtual Artist, which enables users to upload a photo and digitally try on eye shadow, lip color and other beauty products. The app’s newest feature? Color Match, which uses computer vision to select shades of makeup best suited to a customer’s skin tone based on an uploaded photo.

 

4. Creating Art 
A colorful conversation about what constitutes art erupted over GumGum’s Art.ificial experiment, which commissioned five human artists and one robot to create a piece of original art inspired by a collection of 20th-century expressionist works, to see if anyone could tell the difference. Cloudpainter, a machine at the Rutgers University Art and Artificial Intelligence Lab, used a palette of reds, blues and yellows and precisely 13,396 brush strokes to create an image that could easily pass for something made by someone with a pulse. Not to say that AI or computer vision will be taking the place of our own eyes and minds anytime soon—or likely ever. But the remarkably human portrait reminds us of what is possible when human ingenuity joins forces with machines that can see all the colors of the rainbow.

Illustrations by A. Micah Smith