WEDNESDAY 2 AUG 2017 3:08 PM

IMPROVED IMAGE ANALYSIS ON SOCIAL DATA

Machine learning algorithms are being used for different technologies like automated neural machine translation, advanced cyber security systems or data analytics. Research on social networks keeps increasing and leading to innovations like better understanding of markets through image analysis.

 

More than 3 billion images are posted on social media each day but they could not be used as source of information since 85% of them are not accompanied by text that references the brands on them. In order to be able to mine information from images, Crimson Hexagon developed an in-house analytics engine.

Zach Zimmel, director of strategy at Grassriots, says, “Whatever the topic, the language around it is increasingly visual. Before Crimson Hexagon’s image analytics capabilities, we were missing the full picture around what matters, and to whom. Our campaigns, and the causes they support, are the better because of it."

This is a step forward in the marketing industry as the information obtained will help brands and agencies gain more knowledge from photo data on the situations, people and other brands that can be associated with their products. It goes beyond logo identification, it can help customize even more marketing campaigns, product enhancements, identify influencers, among other functions. 

Errol Apostolopoulos, senior vice president of product at Crimson Hexagon, says, “As consumers, especially millennials, continue to create visual content, our system’s machine-learning-based algorithm is well-equipped to ingest, store and classify those images and videos for our clients, so they can get the clearest insights about their customers.”

The software uses a comprehensive image database of more than 125 million images a day to train its machine-learning image analysis algorithm and obtain accurate information. The engine can perform image-specific filtering, data visualizations on posts containing images and a multimedia tile view display of analysed posts.

An example of the capabilities of the engine was establishing by image analysis on Instagram that Coca-Cola and McDonalds are the most commonly paired brands, puppies are the most photographed animal with 37 million puppies per month and 1.4 million visual mentions of avocado toast were posted in 1 month.