Research

My research is on deep learning and computer vision: object detection, segmentation and classification in images. I worked with classical image processing and neural networks between 2010 and 2016 and moved to deep learning after 2018. In recent years my focus has been methodological soundness: patient-level or clip-level data splitting and the effect of the evaluation protocol on reported results.

Biomedical imaging

Developmental dysplasia in hip ultrasound, breast lesions in thermal images, brain tumours in MRI, coronary stenosis in angiography, and tissue perfusion with multi-modal imaging. This work is carried out together with clinicians.

Agriculture and livestock

Body condition scoring, lameness and face recognition in dairy cows; egg fertility detection; ration formulation software. This line has continued since 2012 with animal science colleagues, and in most studies we collect the data ourselves in the field.

Information systems, education and data analytics

Machine learning on educational and organizational data: the effect of distance education content on achievement, mobile learning, text mining, demand and consumption forecasting.

Large language models

I am the academic advisor of an industry project on generating SQL from natural-language questions in Turkish for enterprise software.

Projects

  • Investigation of the Generalizability Performance of Transformer-Based Deep Learning Architectures in the Automatic Segmentation of Coronary Artery Stenosis. Akdeniz University Scientific Research Projects, FBA-2026-7101, 2026 – ongoing, principal investigator.
  • Real-Time Automated Tissue Perfusion Monitoring with Multimodal Sensing. TÜBİTAK 2219 fellowship, University of Nevada, Las Vegas, 2024 – 2025, principal investigator.
  • Computer Aided Egg Fertility Control System. Akdeniz University Scientific Research Projects, FBA-2019-4898, 2019 – 2021, principal investigator.
  • Development of a Textile Products Quality Control System. Techno-enterprise Capital Support, 2014 – 2015, principal investigator.

Computing infrastructure

Models are trained at the National Center for High Performance Computing at Istanbul Technical University (UHeM) and on the NVIDIA DGX Spark system at Akdeniz University.

Collaborators

Ahmet Bozkurt, Hacer Kuduz, Hasan Erdinç Koçer, Mustafa Boğa, Emre Dandıl, Güray Tonguç, and Brendan T. Morris of the University of Nevada, Las Vegas.

Graduate supervision

I supervise master’s and doctoral theses in the Management Information Systems programme of the Institute of Social Sciences at Akdeniz University. Current topics: digital traces of exam behaviour, software developers’ trust in AI tools, the wearable health technology market, pre-screening of gifted students, and the appropriateness of emergency department visits. Prospective students are welcome to e-mail a short CV and the topic they are interested in.

The full list is on the Publications page.