Title:ADA-Net: Attention-Guided Domain Adaptation Network with Contrastive Learning for Standing Dead Tree Segmentation Using Aerial Imagery

Abstract:Information on standing dead trees is important for understanding forest ecosystem functioning and resilience but has been lacking over large geographic regions. Climate change has caused large-scale tree mortality events that can remain undetected due to limited data. In this study, we propose a novel method for segmenting standing dead trees using aerial multispectral orthoimages. Because access to annotated datasets has been a significant problem in forest remote sensing due to the need for forest expertise, we introduce a method for domain transfer by leveraging domain adaptation to learn a transformation from a source domain X to target domain Y. In this Image-to-Image translation task, we aim to utilize available annotations in the target domain by pre-training a segmentation network. When images from a new study site without annotations are introduced (source domain X), these images are transformed into the target domain. Then, transfer learning is applied by inferring the pre-trained network on domain-adapted images. In addition to investigating the feasibility of current domain adaptation approaches for this objective, we propose a novel approach called the Attention-guided Domain Adaptation Network (ADA-Net) with enhanced contrastive learning. Accordingly, the ADA-Net approach provides new state-of-the-art domain adaptation performance levels outperforming existing approaches. We have evaluated the proposed approach using two datasets from Finland and the US. The USA images are converted to the Finland domain, and we show that the synthetic USA2Finland dataset exhibits similar characteristics to the Finland domain images. The software implementation is shared at this https URL. The data is publicly available at this https URL.