How the visual elements of an image can help or hinder audience engagement.
When it comes to visual storytelling, choosing or creating the right image can play an important role in capturing your audience’s attention. An image is not a singular thing, it comprised of thousands of visual elements that come together to create it.
Understanding how these elements effect whether viewers engage with the photo as a whole can help marketers develop a well-founded visual content strategy for social media and blog posts.
This goes beyond just recognizing what visual images do well with an audience. The goal is to break down the images on a micro-level to find the best combination of elements to include in order to receive the most positive feedback from a target audience.
Fortunately, advancements in AI and computer vision make it so we can analyze images on a deeper level than just “a picture of a sunset”. With computer vision we can see and analyze photos for the thousands of design features that make it up and then understand how those features effect performance.
So how does this work?
Computer vision refers to a computer’s ability to “see” images and recognize the various elements of the image. Computer vision works in a similar way to the human eye and brain. It takes in an image and then searches for shapes to differentiate into everyday objects.
The computer is trained through experience. The more objects with descriptors the computer “sees”, the more it is able to understand the characteristics of those objects, and identify them accurately for themselves in the future. This allows a computer to recognize the things that appear in an image such as that the ocean is the ocean, and a boat is a boat.
A computer doesn't look for the core subject, but rather it looks and identifies all of the objects that make it up. By doing this, it breaks an image down to the micro level.
One of the other visual design features that computer vision takes into account is the color pallet of a particular image.
The machine starts by identifying the exact color that make up each pixel in an image, giving it an extremely exact color palette of the image as a whole. However, because many of those colors are indistinguishable from each other to humans, they are then separated into the 12 core color groups and the distinct variations of each of those colors.
The 12 core colors that Cortex recognizes
The defined variations of the color blue
This allows the computer to make both a general and more complex color palette of an image
You are able to see the breakup of both the core 12 colors and the variations of each within a single palette.
We’ve established that an image is more than just a singular image, but an image is also more than just the sum of its parts. How those parts come together is another important element, which is where image composition comes into play.
Not only can AI use machine learning and computer vision to identify the colors and objects that make up an image, but it can also look at the compositional elements like point of view and aspect ratio.
It breaks down and analyzes how photo angles like an aerial vs portrait effect performance. As well as how the width and height of an image effect it.
So why is knowing all the elements of an image and how they come together important?
Because along with all of this creative data, visual content like photographs and infographics on social media come with a lot of performance data. Looking at the two together, machine technology can identify the exact elements in a photo that make it perform well or perform poorly.
This changes the way we think about creating content. Instead of just general suggestions we can understand the exact elements we need to include in a piece of content and the best way to go about combing elements to maximize engagement.
For example, if you are the ski tourism sector of Visit Utah, it helps you understand that while one person skiing down a mountain is good for audience engagement, kids and families skiing are not. More so, if you decide are going to take more photos of single people skiing down a mountain, it can tell you the ideal color pallet your photo should have.
Given a set of photos, Cortex can use the elements that make up images to identify ones that are similar. By cross analyzing over 8,000 visual elements of the different photos with the performance scores they receive on social media it can identify which “clusters” of content perform best, providing a foundation to build a content strategy.
Not only can it compare how one cluster stacks up to another one in terms of audience engagement, but it can also look at how certain elements like color pallet affect the performance of the images within a cluster.