Ref.No: | 61515800 |
Start date: | 17.03.2023 |
End date: | 16.05.2025 |
Approval date: | 10.03.2023 |
Department: | RURAL & SURVEYING ENGINEERING |
Sector: | TOPOGRAPHY SCOPE |
Financier: | 4Η ΠΡΟΚΗΡΥΞΗ ΕΛΙΔΕΚ ΓΙΑ Υ.Δ., ELIDEK |
Budget: | 23.400,00 € |
Public key: | ΨΨΡΨ46ΨΖΣ4-ΖΦΞ |
Scientific Responsible: | Prof. KARANTZALOS |
Email: | karank@central.ntua.gr |
Description: | DESPITE THE SIGNIFICANT NEGATIVE IMPACT OF MARINE POLLUTION ON THE ECOSYSTEM AND HUMANS, ITS AUTOMATED DETECTION AND TRACKING FROM THE CURRENTLY BROADLY AVAILABLE SATELLITE DATA IS STILL A MAJOR CHALLENGE. IN PARTICULAR, MOST RESEARCH AND DEVELOPMENT EFFORTS CONCENTRATE ON MONOLITHIC APPROACHES, IMPLEMENTING TECHNIQUES FOR MAINLY BINARY DETECTION AND CLASSIFICATION TASKS, I.E., DETECT PLASTICS OR NO PLASTICS, DETECT OIL SPILLS OR NOT OIL SPILLS. MOREOVER, THE MAJORITY OF DEVELOPED APPROACHES TEND TO BE LOCAL/REGIONAL OR FAIL TO SCALE AND GENERALIZE ADEQUATELY TOWARDS OPERATIONAL DEPLOYMENTS. THE BLACK-BOX OPERATION OF STATE-OF-THE-ART MACHINE LEARNING AND DEEP LEARNING SOLUTIONS ALSO HINDERS THE UNDERSTANDING OF NEURAL NETWORKS DECISIONS, ASPECTS REGARDING CAUSALITY AND TRANSPARENCY, CONCEALING ANY BIAS AND OTHER SHORTCOMINGS IN MODEL PERFORMANCE. TO THIS END, THIS PHD THESIS AIMS TO TACKLE THE AFOREMENTIONED CHALLENGES BY STUDYING, DESIGNING AND DEVELOPING NOVEL METHODOLOGIES AND EXTENSIVELY VALIDATING OUTCOMES. THE GOAL IS TO STUDY ALL TYPES OF MARINE POLLUTION AND DEVELOP SELF-SUPERVISED, SEMI-SUPERVISED DEEP NEURAL NETWORK APPROACHES UNDER CUTTING-EDGE EXPLAINABILITY FRAMEWORKS. WE ARE TARGETING SCIENTIFIC EXCELLENCE, TOP-RANK INTERNATIONAL CONFERENCES AND HIGH-IMPACT SCIENTIFIC JOURNALS FOR DISSEMINATING PHD OUTCOMES WHILE OPENLY PUBLISHING THE DEVELOPED SOFTWARE AND SOURCE CODE OF THE DEVELOPED METHODS. |