EX-FIQA: Leveraging Intermediate Early eXit Representations from Vision Transformers for Face Image Quality Assessment
IEEE International Conference on Automatic Face and Gesture Recognition, 2026
Curriculum vitae
Ph.D. researcher in face biometrics at Fraunhofer IGD and TU Darmstadt. My work shows that face image quality can be read directly out of the internal computation of a pretrained recognition model, removing the need for quality labels, dedicated training, or added inference cost. Wider interests: efficient and explainable face recognition, presentation attack detection, and foundation models for biometrics.
Experience
Jul 2024 to present
Research Scientist
Fraunhofer IGD · Darmstadt, Germany
- Face image quality assessment, efficient and explainable face recognition, presentation attack detection.
Apr 2023 to Oct 2023
Master's Thesis Student
Siemens AG · Erlangen, Germany
- Hardware aware neural architecture search for semi supervised anomaly detection on microcontrollers. A memory efficient weight sharing NAS pipeline with model based predictors, deployed to STM32L4+.
Jun 2022 to Mar 2023
Working Student, Electronics Engineer
Neura Robotics GmbH · Metzingen, Germany
- MAiRA and MiPA collaborative robot arms.
Feb 2021 to Apr 2021
Machine Learning Intern
Kuartis Technology and Consulting · Ankara, Turkey
- Acoustic anomaly detection with VAEs, GANs, MFCC and HDBSCAN.
Education
Jul 2024 to present
Computer Science, Ph.D.
TU Darmstadt · Darmstadt, Germany
- Robust Face Biometrics, advised by Prof. Dr. Naser Damer.
Sep 2021 to Nov 2023
Machine Learning, M.Sc.
University of Tübingen · Tübingen, Germany
2015 to Jul 2021
Mathematics, B.Sc. and Electrical & Electronics Eng., B.Sc.
Middle East Technical University (ODTÜ) · Ankara, Turkey
- Exchange semester at KAIST, South Korea.
Awards
2026
Best Paper Award
PreFIQs: Face Image Quality Is What Survives Pruning, CVPR 2026 Workshops.
Publications
Full list →14 papers published or accepted at WACV, CVPR, FG, IJCB, ECCV and ACM MM venues. A selection is below.
ATTN-FIQA: Interpretable Attention-based Face Image Quality Assessment with Vision Transformers
IEEE International Conference on Automatic Face and Gesture Recognition, 2026
ViTNT-FIQA: Training-Free Face Image Quality Assessment with Vision Transformers
IEEE/CVF Winter Conference on Applications of Computer Vision Workshops, 2026
FoundPAD: Foundation Models Reloaded for Face Presentation Attack Detection
IEEE/CVF Winter Conference on Applications of Computer Vision Workshops, 2025
Tomatoes, Potatoes, and Onions: Questioning the Need for Faces in Face Presentation Attack Detection
ACM International Conference on Multimedia Workshops, 2026
EXPL-FR: Explaining Face Recognition Models via Vision-Language Alignment
European Conference on Computer Vision Workshops, 2026
Service and community
- •Co organiser, IJCB 2026 Competition on Adapting Foundation Models for Face Recognition Using Synthetic Training Data.
- •Scoutmaster, World Organization of the Scout Movement.
- •Member, AI Grid Initiative.
Skills
Core
PythonPyTorchC/C++MATLABGitDistributed GPU trainingONNXDockerscikit-learnOpenCVNumPySciPypandasSQLMLflowWeights & Biases
Languages
Turkish (native)English (C2)German (B2)