Detecting multiclass objects in remote sensing imagery: A deep learning approach

Abstract

Abstract Object detection from remote sensing imagery is crucial in fields including monitoring agriculture, efficient urban planning, environmental conservation, and disaster response. It enables informed decision-making by identifying and tracking objects such as crops, infrastructure, and environmental changes, contributing to optimized re-source management and rapid response in various domains. Multiclass object detection benefits from the transformative power of deep learning, surpassing traditional methods in accuracy and efficiency. This research investigates the effects of data augmentation techniques and diverse hyperparameters on the training process of a single-stage YOLOv5 deep learning algorithm applied to the selected DIOR remote sensing dataset for multiclass object detection practices, resulting in 0.628 mAP score among the applied experiments.