Spin-out firm in Switzerland uses computer vision to improve recycling purity
WasteFlow has developed an artificial intelligence-based monitoring system that can reportedly identify 66 waste categories.
The start-up uses cameras and sensors to identify, quantify and track waste streams in real time. They claim it can help materials recovery facilities reduce contamination and improve the quality of recycled outputs.
The system is said to be capable of estimating the composition, weight and potential hazards of material moving along sorting lines.
According to the company, classification accuracy reaches 98% for some common packaging polymers, including PET, HDPE and LDPE, with the technology increasing material recovery by an estimated 6%.
Beyond identifying materials, the system is also said to analyse conveyor throughput, bottlenecks and sorting performance to recommend operational adjustments.
Developed in collaboration with EPFL’s Computer Vision Laboratory, the next iteration will seek to integrate infrared, thermal and depth imaging to detect hazardous or partially obscured objects, and ultimately enable automated optimisation of sorting equipment as waste streams change.
