Our site uses cookies. By continuing to browse this site you are agreeing to our use of cookies.  More information
Image Classification through Convolutional Neural Networks
Making creative decisions about your advertising can feel like a crapshoot.

Should you go with what speaks to your emotions, or wait for some data-driven explanation? Does such a thing even exist? This is where image classification can help.

Let’s introduce this new concept by starting with something that most of us are familiar with: our iPhones. Apple changed the game in 2017 with the launch of iPhone X and Face ID. The passwords that were so often forgotten and being reset were instantly replaced by facial recognition software. Through convolution neural networks (CNN) and image classification, our faces have become as unique as the passcodes they have replaced.

Classification of Images

So how does facial recognition work? With an iPhone, the catalog of faces–when provided to the CNN (Convolutional Neural Network) within the device–are simplified to the component pieces and matched with the training data. Each one provides some variation which adds to the final prediction. To understand these distinctions, the CNN must be presented with a set of images to analyze.

As this relates to Face ID and smartphone facial recognition, you’re required to turn the phone in a multitude of angles during setup, teaching it all the contours of your unique face. This process is called “Training”.

Training a Neural Network

Training is another way of saying teaching or allowing to learn, as the network begins to define what will return a positive response. You are slowly helping the CNN operating within your device to get to know you with each additional photo provided. A CNN will attempt to find repeating features within the provided image set. It will then deem these specific features as important in the final classification of whether the person looking into the camera is, in fact, someone that the phone has been told to recognize.

Here’s where it gets interesting: Rather than training with facial structure and expressions, it is possible to train a CNN with any other type of image.

Marketing Advantage of CNNs

Using Image Classification to examine and evaluate creative digital media, we can accomplish a variety of objectives. Whether it be establishing a precise and accurate perspective on the environment our clients are competing in or being able to systematically prove what successful media contains, these are invaluable data points that should inform marketing decisions. To take it a step further, CNNs can even use image classification in social and define what is most frequently effective when it comes to trend development. The possibilities are endless.

At Empower, we’re committed to delivering data-powered creative marketing solutions, and image classification is just that. As the technological landscape continues to grow, so do our capabilities. Data belongs in all places where marketing decisions are being made, including creative ones.