DET: Data Enhancement Technique for Aerial Images
Master thesis
Permanent lenke
https://hdl.handle.net/11250/3020375Utgivelsesdato
2022Metadata
Vis full innførselSamlinger
Sammendrag
Deep learning and computer vision are two thriving research areas within machine learning.In recent years, as the available computing power has grown, it has led to the possibilityof combining the approaches, achieving state-of-the-art results. An area of research thathas greatly benefited from this development is building detection. Although the algorithmsproduce satisfactory results, there are still many limitations. One significant problem is thequality and edge sharpness of the segmentation masks, which are not up to the standardrequired by the mapping industry. The predicted mask boundaries need to be sharper andmore precise to have practical use in map production.
This thesis introduces a novel Data Enhancement Technique (DET) to improve the boundaryquality of segmentation masks. DET has two approaches, Seg-DET, which uses a segmentation network, and Edge-DET, which uses an edge-detection network. Both techniqueshighlight buildings, creating a better input foundation for a secondary segmentation model.Additionally, we introduce ABL(RMI), a new compounding loss consisting of Region MutualInformation Loss (RMI), Lovasz-Softmax Loss (Lovasz), and Active Boundary Loss (ABL).The combination of loss functions in ABL(RMI) is optimized to enhance and improve maskboundaries.
This thesis empirically shows that DET can successfully improve segmentation boundaries,but the practical results suggest that further refinement is needed. Additionally, the results show improvements when using the new compounding loss ABL(RMI) compared to itspredecessor, ABL(CE) which substitutes RMI with Cross-Entropy loss(CE).