Deep Learning for Autonomous Drone Vision

Abstract

Autonomous drones have the potential to reshape numerous industries with the role of advanced drone vision being central to achieving operational autonomy. This thesis marks a significant advancement in autonomous drone vision, tackling key challenges such as data collection, thin structure detection and semantic segmentation.

In response to the pressing need for comprehensive data in this domain, the Drone Depth and Obstacle Segmentation (DDOS) dataset is introduced, specifically designed for drone vision. Using this dataset, a state-of-the-art monocular wire segmentation and depth estimation model is developed to address the challenge of detecting thin structures, which is crucial for the safe flight of autonomous drones.

Another major contribution is the development of recursive denoising, a novel diffusion-based approach to semantic segmentation, which greatly improves scene understanding from aerial perspectives. This enables autonomous drones to better interpret their environment, a critical capability for navigating complex scenarios.

Together, these developments not only propel drone vision technology forward but also advance the broader disciplines of machine learning and computer vision. They showcase the potential of sophisticated data-driven methods to tackle complex real-world challenges, highlighting the evolving capabilities of autonomous drones in understanding and navigating their surroundings effectively.

Related Publications

• Multi-Class Segmentation from Aerial Views using Recursive Noise Diffusion - WACV 2024
• DDOS: The Drone Depth and Obstacle Segmentation Dataset - CVPR (Workshop) 2024
• UCorr: Wire Detection and Depth Estimation for Autonomous Drones - ROBOVIS 2024

Thesis Details

• Author: Benedikt Kolbeinsson
• Institution: Imperial College London
• Department: Department of Electrical and Electronic Engineering
• Supervisor: Krystian Mikolajczyk
• Qualification Name: Doctor of Philosophy (PhD)
• Date Issued: 2024-03-13
• Date Awarded: 2025-02-01
• DOI: 10.25560/116954
• Permanent URI: https://hdl.handle.net/10044/1/116954
• Licence: CC BY-NC 4.0

Keywords

Citation

@phdthesis{kolbeinsson2025thesis,
  author = {Benedikt Kolbeinsson},
  title  = {Deep Learning for Autonomous Drone Vision},
  school = {Imperial College London},
  year   = {2025},
  doi    = {10.25560/116954},
  url    = {https://doi.org/10.25560/116954}
}