My name is Valerijan and I am a computer vision / ML engineer with a strong research background (PhD) working on building robust machine learning systems and non-trivial vision problems under real-world constraints.
My interests lie at the intersection of computer vision, visual attention (eye-tracking), representation learning, and robustness/security of ML systems.
Currently:
- Research collaboration with École Polytechnique (Paris) - dataset ownership verification and poisoning attacks
- PhD track in AI at the Palace of Science, Belgrade
- visual attention & eye-tracking -> Predicting image memorability and recognition outcomes from fixation patterns.
- ** Repo: (https://github.com/ValerijanPro/eye-tracking-image-memorability)
- robust satellite image segmentation -> Adapting SAM-2 for Sentinel-2 and PlanetScope imagery (Serbia region).
- ** Repo: (https://github.com/ValerijanPro/satellite-image-segmentation-advanced)
- 3D vision concepts (AR 3D Unity, medical scans) -> Unity AR visualization, 3D medical image classification (NRRD)
- ** Repos: (https://github.com/ValerijanPro/unity-AR-3D) (https://github.com/ValerijanPro/medical-3d-nrrd-classifcation)
- Panorama reconstruction (OpenCV, C++)
Python • PyTorch/TensorFlow • OpenCV • React • Unity (C#) • Git
- LinkedIn: [(https://www.linkedin.com/in/valerijan-matvejev-17a512226/)]
- Email: [v.matvejev@sbb.rs]