Semi-Supervised Learning for Satellite Maritime Litter Detection
The use of multispectral satellite data, such as Sentinel 2, have already demonstrated useful for the detection of maritime litters [1]. Indeed, AI-based tools such as one presented in [1] allow monitoring maritime areas and identify maritime litters that could be harmful for local faunas and vessels. As a step ahead, through a collaboration with the ESA Φ-lab and potential additional partners, this thesis aims to detect maritime litters with semi-supervised methods, decreasing the needs of time expensive and costly labelled data.
[1] Mifdal, Jamila, Nicolas Longépé, and Marc Rußwurm. Towards detecting floating objects on a global scale with learned spatial features using sentinel 2. No. CONF. 2021.
Space
Research & Development
Smarter Product or Service
Vision
Self/Unsupervised
Image Data