AI for better quality in recycled waste

Important results from the project

The overall goal of the project was to investigate whether AI-based image analysis can be used to detect incorrectly sorted waste at recycling centers. The aim with this to increase the environmental benefit and reduce costs, because if the waste is not sorted correctly, it can not be recycled, but instead goes to incineration or destruction, which is a big waste. The project´s objective has been fulfilled as we have successfully developed an AI that with the help of machine learning through image analysis can detect incorrectly sorted waste in the wooden container and alert our staff.

Expected long term effects

In the project, we have developed an AI that can identify incorrect sortings in the wooden container in about 50% of the cases in which they occur, which from our point of view is a fantastic results as halving unclean material is a huge improvement. AI´s image analysis can automatically detect error sorting and alarms so that our staff can directly pick out the unclean material. The results from the project and the project´s goal fulfillment have exceeded our expectations, and in the project we have also significantly increased our knowledge of AI and insights into its opportunities.

Approach and implementation

The project work has been conducted in five work packages, where each work package has had a defined time period, a specified content and clear deliverables. The project group has consisted of staff from our own organization and researchers from the University of Skövde. The project structure and cooperation between the parties in the project has worked very well and has led to a project that we think is very successful with a results that exceeds our expectations.