Can Rollas

M.Sc. Researcher, Computer Vision

Izmir Institute of Technology (IZTECH)  |  İzmir, Türkiye

canrollas@gmail.comgithub.com/canrollas

I am an M.Sc. researcher in Computer Engineering at IZTECH, supervised by Prof. Dr. Yalın Baştanlar. My thesis focuses on AI-supported wound analysis using 2D and 3D computer vision. My work centers on image segmentation and generative methods for data-scarce medical imaging, with emphasis on generalization across domains and annotation regimes.

Education

Izmir Institute of Technology (IZTECH)
M.Sc. Computer Engineering — Computer Vision
  • Thesis: AI-Supported Wound Analysis Using 2D and 3D Computer Vision
  • Advisor: Prof. Dr. Yalın Baştanlar
  • Medium of instruction: English
Jul. 2024 – Present
İzmir, Türkiye
Izmir Institute of Technology (IZTECH)
B.Sc. Computer Engineering
  • Includes a one-year English preparatory program; completed while working in industry from 2021 onward
  • Medium of instruction: English
  • Thesis: Remote Tracking of Agricultural Fields
Jan. 2018 – Jun. 2024
İzmir, Türkiye

Research Experience

Graduate Researcher
Computer Vision Research Group (CVRG), IZTECH
  • Designed and conducted a systematic wound tissue segmentation benchmark (37 encoder–decoder configurations) forming the basis of an under-review publication.
Jul. 2024 – Present
İzmir, Türkiye
Undergraduate Researcher (B.Sc. Thesis)
IZTECH, Computer Engineering
  • Developed a satellite-based NDVI pipeline for longitudinal agricultural field monitoring with IoT climate station and weather integration.
  • Applied deep-learning-based plant disease detection on field imagery.
Jan. 2023 – Jun. 2024
İzmir, Türkiye

Industry Experience

Machine Learning Engineer
Amatis
  • Designed and deployed deep learning models for melanoma segmentation, wound segmentation, and ECG signal classification on Azure.
  • Built a clinical decision-support chatbot integrating patient data retrieval with a large language model backend.
Aug. 2025 – May 2026
İzmir, Türkiye
Software Developer
Sovrox – MLBNova (Farmavil)
  • Implemented satellite-based NDVI analytics and AI-driven plant disease detection within a cross-platform mobile application.
  • Integrated an LLM-powered agricultural Q&A system and maintained cloud backend infrastructure on AWS.
Jun. 2022 – Jun. 2025
İzmir, Türkiye
Software Developer (part-time)
Somera
  • Built Scrapy-based web crawlers for automated Turkish news collection and deduplication; developed supporting NLP pipelines for text classification and summarization.
Jun. 2021 – Jul. 2022
İzmir, Türkiye

Publications & Preprints

  1. Intra-Wound Tissue Segmentation: A Systematic Comparative Study. Rollas, C., Baştanlar, Y. Accepted from MEDFMB Workshop, ECCV, 2026.
    Systematic benchmark of 37 configurations spanning 9 decoder architectures and 11 encoder backbones on a clinical dataset of 539 annotated wound images; Mix Transformer encoders consistently outperformed CNN families.
  2. Development of an Open-Source Terminal-Based Analog Circuit Simulator. Kersü, A., Rollas, C., Salman, Z.D., Salam, A., Gumus, A. International Conference on Studies in Engineering, Science, and Technology, 2023. [PDF]
    Developed a lightweight open-source circuit simulator for passive components, validated against LTSpice.

Selected Projects

ML-Assisted Melanoma Diagnosis SystemEfficientNet · Grad-CAM · PyTorch
2024

Fine-tuned EfficientNet-B0 on the ISIC dataset (~11k images), reached F1 score of 0.86 on the test set, and integrated Grad-CAM for clinical interpretability.

Mitochondria Segmentation with U-NetU-Net · Keras
2024

Binary semantic segmentation pipeline for electron microscopy images with mean IoU around 0.75.

Technical Skills

Languages:Python, C, C++, SQL, Bash/Shell, JavaScript
ML / DL:PyTorch, TensorFlow/Keras, Scikit-learn, Hugging Face Transformers, Diffusers
Computer Vision:Image Segmentation, Object Detection, Generative Models (GANs, Diffusion), OpenCV
Infrastructure:AWS (Lambda, S3, API Gateway), Azure, Docker, Linux