About me
I am Cansu Korkmaz Soner, a computer vision researcher working on image restoration and generative vision, with a focus on efficient and reliable adaptation of models to real-world degradations. I received my Ph.D. in Electrical and Electronics Engineering from Koç University, where I was advised by Prof. A. Murat Tekalp and Dr. Zafer Dogan. I also received my M.Sc. and B.Sc. degrees in Electrical and Electronics Engineering from Koç University in 2021 and 2019, respectively.
My research spans super-resolution, denoising, deraining, and deblurring, and draws on CNNs, transformers, and diffusion models. I am particularly interested in parameter-efficient learning (e.g., low-rank and adapter-based methods) and frequency-domain modeling using Fourier and wavelet representations. My work has been published at leading venues such as CVPR, ICCV, and ICIP.
I have also conducted research as a Visiting Researcher at the University of Würzburg, working with the Computer Vision Lab under the supervision of Prof. Radu Timofte, focusing on real-world image restoration and efficient model adaptation.
Alongside academia, I work as a Lead AI Engineer, developing privacy-preserving generative vision systems for real-world deployment. My long-term goal is to build trustworthy, efficient, and deployable visual systems, and to pursue an academic career leading independent research and mentoring students.
Beyond research, I am a certified yoga instructor, having completed a 200-hour Yoga Teacher Training (200TT), and an aerial yoga instructor (2025). I maintain an active interest in movement, body awareness, and well-being, which complements my research-driven work. I am also engaged in glass art and have produced several artworks using the glass fusion technique. View my glass art works →
Experience
Lead AI Researcher — Syntonym, Istanbul / London (2025 – Present)
Leading research on privacy-preserving generative vision systems, developing synthetic face and identity-anonymization models for real-world, production deployment.
Visiting Researcher — Computer Vision Lab, University of Würzburg, Germany (2024 – 2026)
Researching real-world image restoration and parameter-efficient model adaptation under the supervision of Prof. Radu Timofte, resulting in AdaptSR and FraIR.
Graduate Research Assistant — Koç University, Electrical and Electronics Engineering (2020 – 2025)
Conducted Ph.D. and M.Sc. research on learned image super-resolution under Prof. A. Murat Tekalp and Dr. Zafer Dogan, spanning multi-model SR, wavelet-domain losses for transformers and GANs, and generative sample selection with diffusion and flow models.
Deep Learning Engineer — Relimetrics, Inc., Berlin, Germany (2019 – 2023)
Built deep learning models for ReliVision, Relimetrics' AI-based visual quality inspection platform, used to automate defect detection and process control on manufacturing production lines.
AI Vision Engineer — HyperbeeAI, Palo Alto, California, USA (2021 – 2022)
Worked on computer vision models, optimizing them to run efficiently on compact, resource-constrained chips.
Software Engineer Intern — Argela Technologies, ITU Ari Teknokent, Istanbul (2018)
Worked on smart factory systems, developing communication protocols for coordinating different products moving through the same production line.
Computer Vision Intern — ISSD, METU Teknokent, Ankara (2017)
Worked on smart traffic solutions, developing classical (non-learning-based) MATLAB algorithms for traffic light detection and OCR.