Institutional Repository
Technical University of Crete
Technical University of Crete
EN
|
EL
| URI: | http://purl.tuc.gr/dl/dias/DAE6AE8C-2DA5-4CBB-A1DC-17EECBCA36F8 | ||
| Year | 2025 | ||
| Type of Item | Diploma Work | ||
| License |
|
||
| Bibliographic Citation | Vasileios Giovanoglou, "Object detection, localization and feature characterization of image data from UAV", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2025 https://doi.org/10.26233/heallink.tuc.103295 | ||
| Appears in Collections |
This thesis explores object detection and feature characterization from visual data collected by unmanned aerial vehicles (UAVs), with a primary focus on leveraging the YOLOv5 model for accurate object recognition and classification in aerial imagery focusing on three classes of objects: cars, buses, pedestrians. The core of the research involves applying YOLOv5 and its variations to the Stanford Drone Dataset, evaluating their performance using various metrics, and conducting experimental studies with different configurations to enhance accuracy and efficiency. The unmodified YOLOv5 model achieved a mAP@.5 89%, outperforming variations, proposed in the thesis, such as YOLOv5 with Softpool and Squeeze-and-Excitation mAP@.5 71% and YOLOv5 with Softpool and CoordAttention mAP@.5 75%. Results highlight the exceptional capability of the YOLOv5x model in object detection, demonstrating its potential for practical applications in surveillance, agriculture, and security. By integrating background isolation techniques and feature detection methods, such as color analysis, this study contributes to the field of computer vision, presenting an advanced approach to UAV-based object detection with significant implications for both academic research and industrial practice.
Copyright © DIAS 2013
