When we developed the tool, we noticed that the application needed to become even more intuitive. It was not always easy to use icons to indicate which cable people have to take, and the color of the cable can vary per customer. People still needed to call Telenet for help, while technical solutions could help them too.

That’s why we decided to implement artificial intelligence (AI) in the tool to create a smart interface. By using AI, the smartphone camera now recognizes any Digicorder cable that the user is holding in his hands. Based on that, the application can show you where to plug in the cable. It’s a simple, effective, and positively transforms the user experience.

‍

We trained our machine learning model by feeding it more than a thousand pictures of each cable it must recognize, and we did tests to verify our model worked with a different set of cable pictures which it had never seen before.

AI is all about making humans and machines work together,  and trust is an essential ingredient for effective collaboration. To that extent, we always try to crack the black-box aspect of machine learning models and make them explain themselves. Here, we explained cable classifications through super-pixels, areas of the image, and color them by their weight of a particular cable type.

For mobile deployment, we use TensorFlow Lite, a subset of TensorFlow which makes it possible to run machine learning models on mobile and embedded devices. Thanks to this technology, we were able to create a fast and compact version of our model that’s only 2.3 MB.

Taken together, we've used AI to recognize electronic cables in real-time, which takes us one step closer towards a great self-service application for installing a Telenet Digibox.