Can AI help us safely dispose dangerous waste?
We all use mobile phones and we bin them again surprisingly often. With a bit of luck, we do so at a collection point and then these end up in sorting centres. There, batteries have to be removed from devices because they contain rare materials and are a fire hazard. Sometimes this succeeds, but sometimes the battery is inside the appliance and the appliance has to be set aside. Human work, because there are too many different devices in circulation, and that means risk of error. Can AI help with this? AI-student Michael Osontuyi looked into the issue.
Unprecedented number of smartphone models
So numerous devices are handed in at recycling points, from those old bricks to brand new smartphones that have experienced an unfortunate fall. There is no database of all devices, Michael and his team concluded, and new devices of all kinds of make are constantly being added. So, can we use an AI to do that analysis? If you let humans do that, there is a risk of distraction or fatigue, causing a battery-powered device to be missed and that could be dangerous.
Core Changemakers
The assignment came from Core Changemakers, an organisation whose TEMNOS system aims to revolutionise waste management with AI and automation ("Shaping a Sustainable Future, Today!"). Several solutions have already been developed for that purpose, which provide accurate object classification with camera and weight sensor. It was precisely in this line that Michael and his team's assignment lay. To enable the system to recognise telephones, a CNN (Convolutional Neural Network) model was used. That is an AI model, which is very good at recognising patterns and so is able to use metadata about the object to signal whether or not there is a battery in the device. So perfect for this question.
Training and applying the model
The challenge was in getting the model working on this specific challenge. This took a lot of work for the team, but ultimately resulted in success. So this is also an application of AI, which can be very valuable in these kinds of processes as an extra check on human actions. The AI is not perfect at catching all devices, but high accuracy was chosen for this application because of the consequences.
Michael tells you more about it in this episode of Eyes on AI.
This project was carried out at the Fontys ICT InnovationLab within the ICT & Artificial Intelligence specialisation. The client is CORE within the Changemakers track, where new, innovative solutions for recycling and waste processing are being developed. Read more on CORE's website.
Meer nieuws
If you ask an ICT professional to think about safety, what will they come up with? An app with an alarm button, or something more ...
If you ask an ICT professional to think about safety, what will they come up with? An app with an alarm button, or something more ...
Two statements on long covid. A: It's between the ears. This is not true; long covid involves demonstrable physical damage. ...