Deep Learning, Machine Learning and their role in waste management

Both Deep Learning and Machine Learning fall under the broader umbrella of Artificial Intelligence, but they differ in how they learn to recognise patterns – a difference that matters a lot when the "pattern" is what a piece of plastic waste looks like.

Differences between Machine Learning and Deep Learning

Machine Learning typically relies on features that are defined in advance – a person decides what characteristics the model should look at. Deep Learning, using neural networks with many layers, can learn relevant features directly from raw data, like images, without those features being specified manually. This makes Deep Learning particularly well suited to visual classification tasks, where the relevant distinguishing features of an object aren't always obvious or easy to define by hand.

Machine Learning and Deep Learning in waste management: the example of Picvisa

PICVISA's equipment uses Deep Learning for the visual classification tasks at the heart of optical sorting – identifying material type, colour and condition from camera images – while other aspects of plant operation may rely on more traditional Machine Learning approaches. The combination allows the system to handle both the visual complexity of sorting and the more structured analytical tasks that support overall plant optimisation.

Explore our solutions