We are advertising several thesis positions, please see below.
Best regards,
Guray
---------------------------------------------------------------------------------------------------------------------
Contact@Fraunhofer IGD: Guray Ozgur (guray.ozgur(a)igd.fraunhofer.de<mailto:guray.ozgur@igd.fraunhofer.de>). Competence Center Smart Living and Biometric Technologies, Fraunhofer IGD
Motivation:
Biometrics is a rapidly growing technology that aims to identify or verify people's identities based on their physical or behavioral properties. Different aspects of biometric technology are active research fields. Enhancing the accuracy of biometric comparisons, securing the biometric templates, managing fast searches in biometric databases, and detecting different attacks on biometric systems, are all essential advancements to enable a wider and more secure deployment of the technology. Most biometric systems are based on image analyses. Therefore, exciting challenges in the computer vision domain are inherited by biometric systems. Such challenges are related to miniature deep learning networks, explainable decisions, and domain adaption.
Our team is offering several open thesis positions to tackle these challenges. Interested students from Computer Science, Data Science, Physics, Statistics, and Mathematics are encouraged to apply. The exact details of the thesis topic can be built on the available topics and the competencies and interests of the student.
The thesis:
The goal of the thesis is to perform state-of-the-art research in computer vision and machine learning for biometrics applications. The exact topic description can be tailored based on the research direction and student interests. The topics can target one of the following domains:
* Robust Face Recognition: Explore novel methods to enhance face recognition performance in challenging conditions, such as varying lighting, pose, and occlusion. Develop strategies to train robust biometric systems in the presence of noisy or incomplete data, improving real-world performance.
* Knowledge Distillation: Investigate techniques to compress complex biometric models while preserving accuracy, enabling efficient deployment on resource-constrained devices.
* Uncertainty Quantification: Focus on methods to estimate and represent the uncertainty associated with biometric predictions, leading to more reliable systems.
* Explainability: Work on developing explainable AI techniques for biometric systems, enhancing transparency and trust in these critical technologies.
* Fairness: Address the critical issue of fairness in biometric systems by developing methods to mitigate bias and ensure equitable performance across different demographic groups.
Required skills: Interest in machine learning and computer vision, good programming skills (Python).
Study programs: Computer Science, Data Science, Physics, Statistics, and Mathematics
Contact: Guray Ozgur (guray.ozgur(a)igd.fraunhofer.de<mailto:guray.ozgur@igd.fraunhofer.de>)
Key literature:
[1] Boutros, Fadi, et al. "Elasticface: Elastic margin loss for deep face recognition." Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2022.
[2] Boutros, Fadi, et al. "CR-FIQA: face image quality assessment by learning sample relative classifiability." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2023.
[3] Boutros, Fadi, et al. "Idiff-face: Synthetic-based face recognition through fizzy identity-conditioned diffusion model." Proceedings of the IEEE/CVF International Conference on Computer Vision. 2023.
[4] Damer, Naser, et al. "Privacy-friendly synthetic data for the development of face morphing attack detectors." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2022.
[5] Kolf, Jan Niklas, et al. "Identity-driven three-player generative adversarial network for synthetic-based face recognition." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2023.
---------------------------------------------------------------------------------------------------------------------
