The research team successfully developed an AI-based detector that more accurately identifies floating plastics in the ocean using satellite images. Utilizing technologies such as deep neural networks, the detector provides more precise location data for cleaning up ocean plastics. By combining Sentinel-2 and PlanetScope images, it enables continuous monitoring of marine debris, offering dual views to improve drift direction models. Even under harsh conditions, such as cloud cover, the model can still accurately identify ocean debris, carrying significant environmental importance.

Scientists have successfully trained an artificial intelligence system to map massive icebergs using satellite images, achieving speeds 10,000 times faster than humans. The U-net algorithm was employed to distinguish icebergs from other features, enabling accurate mapping of iceberg surface area and contours in one hundredth of a second. Iceberg monitoring is crucial for maritime safety and scientific research, and this technology promises to provide new insights into iceberg shapes and sizes. The study indicates that machine learning will enable scientists to monitor remote areas almost in real time, providing a powerful tool for polar environmental research.