Master's Thesis (f/m/d) Drone Tracking

Deutsches Zentrum für Luft- und Raumfahrt e.V. · Sankt Augustin

ThesisGerman needed: not said

Still open on 1 Nov

You apply on the company's own website.

Key facts

Familiarization with the state of the art in camera-based drone detection as well as multi- and single-object tracking (in particular tracking-by-detection methods and transformer-based approaches)
Selection, preparation and, if necessary, annotation of suitable video data sets as well as definition of realistic evaluation scenarios including occlusion and interfering objects
Design and implementation of a modular "detect-then-track" pipeline that builds on initial detections and continues tracking independently
Integration and comparative evaluation of representative tracking methods of both paradigms
Systematic evaluation based on established metrics as well as targeted investigations on the influence of initial detections
Critical analysis and classification of results with regard to suitability for real surveillance deployment
Documentation of methodology and results in a written thesis

Ongoing master's degree in computer science, mathematics, optical engineering or a comparable course of study
Good programming skills in Python as well as practical experience with deep learning frameworks (ideally PyTorch)
Basic understanding of computer vision and deep learning
Independent and structured way of working

DLR stands for diversity, appreciation and equality for all people. We promote independent work and the individual development of our employees in their personal and professional environment. For this purpose, our numerous training and further education opportunities are available to you. Equal opportunities are a particular concern for us; we therefore want to increase the proportion of women in science and management in particular. Applications from severely disabled persons are given preference in case of professional suitability.

Familiarization with the state of the art in camera-based drone detection as well as multi- and single-object tracking (in particular tracking-by-detection methods and transformer-based approaches)
Selection, preparation and, if necessary, annotation of suitable video data sets as well as definition of realistic evaluation scenarios including occlusion and interfering objects
Design and implementation of a modular "detect-then-track" pipeline that builds on initial detections and continues tracking independently
Integration and comparative evaluation of representative tracking methods of both paradigms
Systematic evaluation based on established metrics as well as targeted investigations on the influence of initial detections
Critical analysis and classification of results with regard to suitability for real surveillance deployment
Documentation of methodology and results in a written thesis

Ongoing master's degree in computer science, mathematics, optical engineering or a comparable course of study
Good programming skills in Python as well as practical experience with deep learning frameworks (ideally PyTorch)
Basic understanding of computer vision and deep learning
Independent and structured way of working

DLR stands for diversity, appreciation and equality for all people. We promote independent work and the individual development of our employees in their personal and professional environment. For this purpose, our numerous training and further education opportunities are available to you. Equal opportunities are a particular concern for us; we therefore want to increase the proportion of women in science and management in particular. Applications from severely disabled persons are given preference in case of professional suitability.