Image recognition, robotisation and AI in recycling
29 November 201811:45 - 12:00
If we want robots to sort, we need to teach them to see. The Recycleye algorithm is a recently developed machine vision system that identifies and classifies waste items based on simple colour images. The neural network can classify items; by material (paper, glass, cardboard, metal and plastic) with up to 97.5% accuracy, and by brand (e.g. a Coca-cola can vs a Pepsi can) with up to 98.2% accuracy. The system’s ability to ‘label’ every single item in any environment combined with its very low cost (approx. 100$ per module), renders the ubiquitous item-level tracking of waste possible. Recycleye is working on using the technology to allow extended producer responsibility schemes to move towards individualization, MRF facilities to meet the growing requirements for recyclate purity by combining the system with a robotic arm, enable local & on street robotic recycling and develop a swarm of autonomous cleaning robots.