Spot on Waste - Using AI to detect litter in the public space
As we all know, littering is a huge problem that is facing governments and citizens alike. Not only does it impact our quality of living, but it also impacts public health. Through plastic litter (a majority of litter in public spaces) micro plastics end up in the soil, water and air.
Currently, the Province of Utrecht (programme Innovation Healthy Urban Living) is undertaking a project in which we develop real-time mapping of litter in public spaces, by using a mobile phone app that automatically takes pictures, while users are walking, cycling or driving. These pictures are anonymized, and afterwards analyzed by an algorithm that is trained to detect litter.
The output is then converted to a publicly accessible heatmap, which indicates per geographical area where litter is an actual problem. This data can be used by trash collectors to optimize their route planning, and can be used to stimulate citizens to gain a stronger sense of duty regarding the cleaning of their own neighborhoods.
The technical development has been done by Camenai, a start-up selected by the Province of Utrecht through a public challenge competition. The Challenge competition was set up to find practical use cases for citizens science, leading to collaboration between citizen scientists and local governments. Their experiences and needs are further taken up by Camenai to translate them into better functionalities in what the algorithm detects, and what the data dashboards display. As such, the project is a triple helix collaboration.
Outputs that have been delivered so far are an increased awareness of litter issues by citizens and roadmanagers. Additionally, we are experimenting with different whays how to display the data in a way that it actually helps citizens and trash collectors in optimizing their initaitives and assignments.
We are developing the solution in an iterative process, and we are discovering several ways in which the solution can be scaled up.
Within the AI solution, we use random samples to examine the effectiveness of the algorithm, which now yields an accuracy above 90 percent, and is still increasing.
Problem or opportunity
Combatting the problem of microplastics in the environment through littering, faced by countries and local governments, and increasing public awareness and desire to help address the problem of littering.
Cities and provences get a clearer indication how big the problem is, and where most litter accumulates. Trash collectors (volunteers and professionals) can change their logistics in a data-driven way, to optimize their impact. In adition, other policies can be employed to counter litter in general or at specific spots.
Through an mobile phone/tablet app, the mobile device's camera makes pictures and videos, uploads them to a server. Then, the images are made anonimous with face and car registration blurring technology. Then, the original images are destroyed to avoid privacy violations. With artificial intelligence, the images are analysed for waste/litter, detecting different categories of litter. These analyses is based on different algoritms which are self-learning. By visualising the litter on a map, the problem is made more tangible.