Semantic segmentation of high-resolution Aerial imagery using an attention-guided Deep Neural Network

The investigation of Land cover surface is an important task for many years, and the evolution of remote sensing technologies has transformed this task.Numerous methods have been designed by researchers to classify the land cover into various classes due to its importance in fields like urbanization, disaster management, transportation, etc.In this work, an attention-guided deep neural network based on encoder encoder-decoder architecture is designed to segment high-resolution aerial imagery into different land cover classes.The proposed method is computationally efficient due to the introduction of an attention mechanism, which makes the model computationally efficient as compared to counterpart methods.The performance of the proposed method is evaluated on evaluation metrics like precision, recall, accuracy and f1-score.The proposed method has obtained an overall accuracy of 71% and has shown satisfactory performance in various classes.