Accuracy, Efficiency, and Transferability of a Deep Learning Model for Mapping Retrogressive Thaw Slumps across the Canadian Arctic

Deep learning has been used for mapping retrogressive thaw slumps and other periglacial landforms but its application is still limited to local study areas. To understand the accuracy, efficiency, transferability of a deep model (i.e., DeepLabv3+) when applied large areas or multiple regions, we conducted several experiments using training data from three different regions across Canadian Arctic. overcome main challenge transferability, generative adversarial network (GAN) called CycleGAN produce new in an attempt improve transferability. The results show that (1) augmentation can accuracy does not guarantee (2) it necessary choose good combination hyper-parameters (e.g., backbones rate) achieve optimal trade-off between (3) GAN significantly if variation source target dominated by color general texture. Our suggest future should prioritize collection where cannot