SegForestNet: Spatial-Partitioning-Based Aerial Image Segmentation

arXiv:2302.01585v2 Announce Type: replace
Abstract: Aerial image segmentation is the basis for applications such as automatically creating maps or tracking deforestation. In true orthophotos, which are often used in these applications, many objects and regions can be approximated well by polygons. However, this fact is rarely exploited by state-of-the-art semantic segmentation models. Instead, most models allow unnecessary degrees of freedom in their predictions by allowing arbitrary region shapes. We therefore present a refinement of our deep learning model which predicts …

abstract aerial applications art arxiv cs.cv deforestation image maps objects partitioning segmentation semantic spatial state tracking true type