Automated identification of mis-sorted household waste

Important results from the project

The project took Sortrace from prototype to operational product. The AI identifies missorted household waste directly during emptying and links each detection to the correct city district. An anonymization module ensures GDPR compliance before images are stored. The municipalities follow the results in a web portal with image evidence, trends and map view. The system detects over 20 different types of materials.

Expected long term effects

Municipalities receive ongoing data on sorting quality and can target information, training and fees where they are most beneficial instead of relying on general campaigns. In the long term, this will result in higher material recycling, less waste for incineration and lower costs, and thus a concrete tool for reaching the EU´s objective of 55 percent material recycling by 2025 and 65 percent by 2035. The solution is directly transferable to other Nordic and European markets.

Approach and implementation

The project was implemented in four work packages: further development of AI and platform, installation on more vehicles in the pilot municipalities, integration with the municipalities´ systems, and productization prior to rollout to new municipalities. The work was conducted iteratively together with the pilot municipalities, with ongoing reconciliations that allowed eligibility requirements and priorities to be adjusted along the way.