Bilkent University
Department of Computer Engineering

---

CS 590/690 SEMINAR

 

Contextual Object Detection in Remote Sensing Images

 

Vahid Namakshenas
Ph.D. Student
(Supervisor: Prof.Dr.Selim Aksoy)
Computer Engineering Department
Bilkent University

Abstract: Object detection in optical remote sensing images is crucial for applications such as geospatial analysis, urban planning, and military operations. While neural networks have significantly improved the precision of object detection, they sometimes mislabel objects in an image. To address this challenge, our study leverages contextual information to enhance prediction accuracy. We introduce a method that combines Conditional Random Fields (CRFs) with a self-attention mechanism. This fusion allows our model to more effectively utilize spatial information, improving its ability to accurately label objects based on their spatial relationships. Our approach aims to refine the accuracy of object detection, ensuring that the identified labels more accurately reflect the objects present in the images. By doing so, we enhance the reliability of object detection in critical applications where exactness is paramount, without adding undue complexity to the model.

 

DATE: March 25, Monday @ 13:30 Place: EA 502