10:45 〜 11:00
[MIS02-07] botell.ai: Open Source Plastic Bottle Tracking
★Invited Papers
キーワード：Plastic Bottles, Computer Vision, AI, Waste, Microplastics, Macroplastics
Plastic pollution poses a stubborn threat to landscapes, freshwater systems, and marine ecosystems worldwide. Plastic debris often originates on land and is transported through watersheds to the ocean, breaking down into micro- and nanoplastics that harm aquatic organisms. Identifying the source of macroplastic pollution is a critical step in addressing this crisis. To support these efforts, the AnthroHydro Lab at The Catholic University of America developed botell.ai, an innovative, open-source tool that leverages computer vision and machine learning to track plastic bottles as they float downstream. The research team trained a YOLOv8 model integrated with NORFAIR, incorporating background removal and advanced post-processing techniques to enhance accuracy. The model consistently achieves precision and recall scores exceeding 0.95 and can operate from various vantage points, including streambanks, bridges, and drones. The model was developed using the Anaconda Python environment, botell.ai is available for public use on GitHub, offering a scalable solution for monitoring and mitigating plastic pollution in waterways.