Publication

Abstract : Object detection within images is a challenging problem in the field of computer vision. This dilemma is interesting when it comes to aerial images, exacerbated by the visual appearances and sizable field of view. Despite a plethora of techniques reported in literature, for object detection in aerial images, a comprehensive survey of the pertinent state-of-the-art techniques is grossly amiss in the computer vision literature. This paper presents an insightful review of the reported methods and techniques deployed towards the task of object detection in aerial images. This survey is for generic object detection in aerial images, unrestricted to any specific class of objects. Pursuant to a review of around 170 research papers, the relevant techniques were broadly categorized into Convolutional Neural Network (CNN), Graph Neural Network (GNN), and Transformer based techniques. The methodology adopted for challenging scenarios such as detection of small objects, multiple objects of different scales and salient object detection is discussed in this article. Besides, details gleaned from publicly available datasets and the standard evaluation metrics germane to detection of aerial objects have been included. The paper ends with the discussion of future research directions in the field to enable the researchers to work on.