Dissertation defence (Computer Science): DI Luca Zelioli
Time
6.9.2024 at 12.00 - 16.00
DI Luca Zelioli defends the dissertation in Computer Science titled ”Leveraging Machine Learning for Maritime Object Detection and Peatland Classification” at the University of Turku on 6 September 2024 at 12.00 (University of Turku, Agora, XXII auditorium, Vesilinnantie 3, 20500 Turku).
Doctoral Dissertation at UTUPub: https://urn.fi/URN:NBN:fi-fe2024080663810
Opponent: Professor Heikki Kälviäinen, LUT University
Custos: Professor Jukka Heikkonen, University of Turku
***
Summary of the Doctoral Dissertation:
Artificial intelligence is present in our daily lives, driving innovation across various sectors, enhancing decision-making processes, and automating complex tasks. This dissertation focuses on two application sectors machine learning plays a central role. The first sector is the maritime environment, it empowers situational awareness models around maritime vehicles. The second sector is remote sensing, with a particular focus on a soil-type classification approach.
The maritime studies proposed in the dissertation investigate how traditional Convolutional Neural Network (CNN) architectures perform in detecting and recognizing objects in the environment. Additionally, Transfer Learning (TL) is implemented to enhance the quality of the detection process. Furthermore, a precisely manually annotated dataset is developed, providing a solid basis for the development of efficient object detection and tracking.
The remote sensing section of deals with the development and evaluation of a methodology that that begins with Geographic Information System (GIS) data inputs and culminates in the creation of a soil-type classification map, with a particular focus on pixel-wise soil-type classification. The proposed peatland methodologies summarize the accumulation of decayed material, aiding in the planning of agricultural or urban areas.
In conclusion, this dissertation demonstrates the role of artificial intelligence in both maritime and remote sensing applications. Advanced CNN architectures and TL enhance object detection in maritime environments, while the developed GIS-based methodology for soil-type classification aids in agricultural and urban planning. These innovations highlight the transformative impact of AI in improving efficiency and decision-making across various sectors.
Doctoral Dissertation at UTUPub: https://urn.fi/URN:NBN:fi-fe2024080663810
Opponent: Professor Heikki Kälviäinen, LUT University
Custos: Professor Jukka Heikkonen, University of Turku
***
Summary of the Doctoral Dissertation:
Artificial intelligence is present in our daily lives, driving innovation across various sectors, enhancing decision-making processes, and automating complex tasks. This dissertation focuses on two application sectors machine learning plays a central role. The first sector is the maritime environment, it empowers situational awareness models around maritime vehicles. The second sector is remote sensing, with a particular focus on a soil-type classification approach.
The maritime studies proposed in the dissertation investigate how traditional Convolutional Neural Network (CNN) architectures perform in detecting and recognizing objects in the environment. Additionally, Transfer Learning (TL) is implemented to enhance the quality of the detection process. Furthermore, a precisely manually annotated dataset is developed, providing a solid basis for the development of efficient object detection and tracking.
The remote sensing section of deals with the development and evaluation of a methodology that that begins with Geographic Information System (GIS) data inputs and culminates in the creation of a soil-type classification map, with a particular focus on pixel-wise soil-type classification. The proposed peatland methodologies summarize the accumulation of decayed material, aiding in the planning of agricultural or urban areas.
In conclusion, this dissertation demonstrates the role of artificial intelligence in both maritime and remote sensing applications. Advanced CNN architectures and TL enhance object detection in maritime environments, while the developed GIS-based methodology for soil-type classification aids in agricultural and urban planning. These innovations highlight the transformative impact of AI in improving efficiency and decision-making across various sectors.
viestintä