Department of Computer Engineering
OBJECT DETECTION IN REMOTE SENSING
(Supervisor: Prof. Dr. Selim Aksoy)
Computer Engineering Department
Object detection has been a crucial part of the computer vision field that attracts a credible amount of interest with different applications in the real world. In recent years, with the advances that emerged in convolutional neural networks, efforts on object detection in natural scenes have been accelerated greatly with the increase in accuracy and robustness of detected objects. Whereas accuracy and robustness of object detection in aerial imagery still fall behind the recent developments of object detection in natural scenes because of the indigenous properties of objects appearing in aerial imagery with arbitrary rotations, varying scales and the unnormalized number of instances in each image. In terms of these aspects, object detection in aerial imagery is still a challenging task in the field of computer vision. In this presentation, we discuss the most recent work in the literature including a novel network for object detection in aerial imagery that exploits the local and global context of objects as well as attention-modulated features. The network learns correlations between global and local contexts of objects. It includes a spatial and scale-aware attention module that learns features of different scales and guides the network to more informative areas where there is possibly an object in that area. We will conclude with open problems and future research directions.
DATE: 26 April 2021, Monday @ 16:00