Assessing the Use of AI to Quantify Plastic Pollution in Rivers

The Surfrider Foundation has explored the use of artificial intelligence to map and quantify plastic pollution in European rivers.  Its aim is to empower citizens to contribute to data collection efforts by using the smartphone app Plastic Origins to film riverbanks during kayaking excursions. The AI model developed, named Surfnet, analyses these videos to detect macro-waste by combining computer vision technologies with video tracking. The use of AI enables better standardised and larger-scale data collection as compared to manual methods, which are often prone to observational errors. A training dataset of 5,000 labelled images was created with the help of volunteers, and techniques such as data augmentation were used to enhance performance. The project also highlights a frugal approach by using TinyML technologies to minimise both the financial and environmental costs of AI, particularly by reducing dependency on cloud-based operations.  While the project effectively raises awareness about plastic pollution among citizens and decision-makers, challenges remain, both in the field and technologically, to make it a true diagnostic tool for plastic pollution in rivers. The entirety of the data generated over the course of this project is available as open data to enable further testing and adaptation for diverse objectives. This open and collaborative approach, relatively rare in AI projects, promotes transparency and encourages the appropriation of the technology for more sustainable innovation.