One Stage Deep Learning Based Method for Agricultural Parcel Boundary Delineation in Satellite Images

Boundary delineation is a rapidly evolving, significant research issue. It can be defined as identifying individual agricultural parcels and accurately outline their borders. Boundary delineation plays an important role in various application related to smart agriculture and precision farming. In this paper, a one stage deep learning method is utilized for boundary delineation. This is done by performing transfer learning and fine tuning the network using thousands of parcels. Later, the result of this network is compared to state-of-the-art models using spatially/temporally different dataset. The results are later discussed using common metrics.