Marmara University

Intelligent Systems Research Lab

Projects

Semi-Supervised Federated Learning Ongoing

Summary: The TUBITAK 1001 Project (2022–2025), led by Prof. Çiğdem Eroğlu Erdem with researcher Assoc. Prof. Ömer Korçak, focuses on enhancing semi-supervised federated learning (SSFL) to improve model performance despite limited labeled data. This project introduces a novel framework combining curriculum and self-paced learning in SSFL, a first in the field, to systematically handle labeled and unlabeled data. Additionally, it proposes a new labeling approach for low-confidence data and an innovative model aggregation method, aiming to address challenges in SSFL and boost learning efficiency in decentralized environments.

Learn more about the Project Here
Self-Supervised Federated Learning for Face Presentation Attack Detection Ongoing

Summary: The TUBITAK 1002 Project (2024–2025), led by Mehmet Fatih Gündoğar with consultants Prof. Çiğdem Eroğlu Erdem and Assoc. Prof. Ömer Korçak, seeks to improve facial recognition systems' resilience against presentation attacks, such as spoofing with printed images or realistic masks. This project introduces a novel self-supervised federated learning (FL) framework for presentation attack detection (PAD), addressing limitations in existing datasets and domain shift challenges. By enabling decentralized, self-supervised training on diverse, unlabeled local data, this approach aims to enhance the reliability and adaptability of deep learning models in PAD without centralizing sensitive personal data.

Learn more about the Project Here
Soft Biometric Data Extraction from Spontaneous Multimodal Affective Face Videos Completed

Summary: The TUBITAK 1001 Project (2017–2020), led by Prof. Dr. Çiğdem Eroğlu Erdem, aims to enhance static texture-based face recognition systems using soft biometrics. By incorporating person-specific facial dynamics during expressions or speech, the project seeks to improve recognition reliability under challenging conditions, such as low resolution or spoofing attacks. This dynamic "signature" will be fused with traditional texture data for more robust biometric identification.

Learn more about the Project Here
Energy-Aware Combinatorial Resource Allocation and Scheduling Models for Cloud Computing Completed

Summary: The TUBITAK 3501 Project (2019–2022), led by Asst. Prof. Ali Haydar Özer, focuses on reducing the energy usage and carbon footprint of cloud computing infrastructures. Through energy-aware resource allocation and scheduling models, users can submit complex resource requests using a specialized language, enabling efficient resource handling and priority setting for dependent tasks. The project aims to support sustainable cloud services, addressing environmental concerns associated with high-energy cloud operations.

Learn more about the Project Here
Incentive Mechanisms for User Provided Networks Completed

Summary: The TUBITAK 3501 Project (2015–2019), led by Asst. Prof. Ömer Korçak, focuses on creating incentive mechanisms for User Provider Networks (UPNs), where users share their cellular connections with those lacking access. The project explores energy consumption, data costs, and user demands, applying game theory to design fair and effective incentives. The goal is to enhance efficiency, fairness, and trust in both self-organizing and operator-controlled UPNs.

Learn more about the Project Here

Thesis

Ongoing
  1. Kübra Uludağ (supervisors: Ömer Korçak, Çiğdem Eroğlu Erdem), Ph.D.
  2. İlhan Demirci (supervisor: Ömer Korçak), Ph.D.
  3. Mehmet Fatih Gündoğar (supervisors: Çiğdem Eroğlu Erdem, Ömer Korçak), Ph.D.
  4. Yunus Emre Alçakakan (supervisors: Ömer Korçak, Çiğdem Eroğlu Erdem), Ph.D.
  5. Okan Çandır (supervisor: Ali Haydar Özer), Ph.D.
  6. Fatma Zehra Güç (supervisor: Ali Haydar Özer), M.Sc.
  7. Merve Günaydın Karakaş (supervisor: Ömer Korçak), M.Sc.
Completed
  1. Sinan Emre Taşçi (supervisor: Ömer Korçak), “Incentive Mechanisms for User Provided Networks”, Ph.D. Thesis, Marmara University, Istanbul, 2023.
  2. Halil Demirezen (supervisor: Çiğdem Eroğlu Erdem), “Remote Heart Rate Estimation Using Non-Contact Photoplethysmography”, Ph.D. Thesis, Marmara University, Istanbul, 2023.
  3. Murat Taşkıran (co-supervised by Çiğdem Eroğlu Erdem), “Hybrid Face Recognition Under Adverse Conditions Using Appearance-based and Dynamic Features of Facial Expressions”, Ph.D. Thesis, Yıldız Technical University, İstanbul, January 2022.
  4. Ayşenur Uslu (supervisor: Ali Haydar Özer), “A Multidimensional Resource Allocation Model for Cloud Computing”, MSc Thesis, Marmara University, İstanbul, March 2024.
  5. Deviliers Caleb Dube, Boğaziçi University, 2024.
  6. Ömer Kaan Karahan (supervisor: Çiğdem Eroğlu Erdem), “Prediction of the Length of Stay and Readmission Probability of Intensive Care Unit Patients”, M.Sc. Thesis, Marmara University, İstanbul, March 2023.
  7. Erbil Öner (supervisor: Ali Haydar Özer), Energy-Aware Combinatorial Resource Scheduling Model for Cloud Computing, MSc Thesis, Marmara University, Istanbul, 2022.
  8. Candar Karabulut (supervisor: Ömer Korçak), S-AIS Based Maritime Network: Distributed Vessel Clustering, MSc Thesis, Marmara University, Istanbul, 2022.
  9. Taj Saleh (supervisor: Ömer Korçak), Comparison of the Effects of Data Privacy Preserving Methods on Machine Learning Algorithms in IoT, MSc Thesis, Marmara University, Istanbul, 2022.
  10. Zeynep Nur Saraçbaşı (supervisor: Çiğdem Eroğlu Erdem), “Face Recognition Using Facial Dynamics of Emotional Expressions”, M.Sc. Thesis, Marmara University, İstanbul, October 2021.
  11. Ali Bolu (supervisor: Ömer Korçak), Path and Task Planning for Multiple Mobile Robots in Smart Warehouse, MSc Thesis, Marmara University, Istanbul, 2021.
  12. Barış Büyüktaş (co-supervised by Çiğdem Eroğlu Erdem), “Curriculum and Active Self-Paced Learning with Minimum Sparse Reconstruction for Face Recognition”, M.Sc. Thesis, Özyeğin University, İstanbul, August 2020.
  13. Galip Pala (supervisor: Çiğdem Eroğlu Erdem), “A Comparative Study of Deep Learning Based Face Recognition Algorithms for Video Under Adverse Conditions”, M.Sc. Thesis, Marmara University, İstanbul, July 2019.
  14. Mustafa Gamsız (supervisor: Ali Haydar Özer), An Energy-Aware Combinatorial Virtual Machine Allocation Model for Green Cloud Computing, MSc Thesis, Marmara University, Istanbul, 2019.
  15. Kenan Cebeci (supervisor: Ömer Korçak), Design of a Queue-Based Microservices Architecture and Performance Comparison with Monolith Architecture, MSc Thesis, Marmara University, Istanbul, 2019.
  16. Kübra Uludağ (supervisor: Ömer Korçak), Energy and Data Rate Modeling in User Provided Networks with Incentive Mechanisms, MSc Thesis, Marmara University, Istanbul, 2018.
  17. Kerem Aytaç (supervisor: Ömer Korçak), Internet of Things-Based Waste and Service Management in Quick Service Restaurants, MSc Thesis, Marmara University, Istanbul, 2018.
  18. İlhan Demirci (supervisor: Ömer Korçak), Gap-Free Load Balancing in Wireless LAN Networks Using Cell Breathing Technique, MSc Thesis, Marmara University, Istanbul, 2014.

Industry Collaborations

  1. “Fast Routing of Profinet Packets in Real-Time Industrial Communication Systems”, Siemens, Academic consultant (Assoc. Prof. Ömer Korçak), 2024 – cont.
  2. “Scoring of Candidate Patients using Machine Learning for Health Tourism”, KOSGEB project, Academic Consultant (Prof. Çiğdem Eroğlu Erdem), 2022-2023.
  3. “Research and Development of Multi-Sided Smart Electronics Market Model Design and Related Optimization Solution Methods”, NeoSwap, USA, June 2022 – cont. (Assoc. Prof. Ali Haydar Özer is the Head of Research in NeoSwap in the context of this Project.)
  4. “BeeSmart Intelligent Warehouse Management System”, Tekhnelogos, Academic consultant (Assoc. Prof. Ömer Korçak), 2021.
  5. In addition to these projects, various R&D studies related to IoT intelligence have been conducted with Koç Digital, and joint papers have been published.