ABOUT

Recycling is an important step towards a circular economy for plastic. In order to effectively recycle plastic materials, it is first necessary to sort the mixed plastic solid waste by polymer type and possible additives (especially dyes). This process is complicated, expensive, and often provides inaccurate results, hence insufficient for further recycling steps.
We propose an automated optical sorting service that uses artificial intelligence (AI) and hyperspectral (HS) data to discriminate between different plastic types. In order to train our deep learning model, we will collect – and make available – a labelled dataset of hyperspectral images (in the visible and near-infrared (VNIR) regions of the spectrum) of different types of plastic solid waste.
The objective is to release a model-as-a-service that can be easily integrated on the materials recovery facilities (MRFs) conveyor belts, providing the basis for an automated, high-speed and accurate sorting system for plastic waste.