Floating marine plastic debris detection with a convolutional neural network and using multispectral satellite images

My PhD thesis aims to improve our understanding of the fate of floating plastic debris, its distribution and sources, as well as the increasing threat posed by harmful foam following Phaeocystis globosa blooms. The project proposes an innovative approach that exploits the vast amount of data available today. A convolutional neural network with the architecture of a U-Net is trained with Sentinel-2 data . Quantifying their present and past presence will provide a better understanding of the current situation and future scenarios. Ultimately, it will contribute to the scientific basis for an approach to reducing the amount of plastic released into the oceans, and a better understanding of the evolution of plastic debris and foam.