Lead: Anna Papp (Columbia University)
Advisor: Amir Jina (University of Chicago)
Manually collecting and counting marine plastics is impractical, which makes it difficult to quantify effects of environmental policy. To mitigate this, we propose a system that is capable of detecting marine debris using satellite imagery. We deploy a two-stage approach for classification. First is at the pixel level to discretize the image, and second at the shape level separating non-water pixels with water pixels. Our system is scalable such that it can generate time-series data for multi-year and multiple basins.
Contributions: Large Scale Computing, Computer Vision