Osman Ulger

Logo

Machine Learning/Computer Vision Engineer | PhD Graduate in Computer Vision at University of Amsterdam

View My GitHub Profile

ATLAS lab logo

Hello!

I am a PhD graduate from the ATLAS Lab, a collaboration between the University of Amsterdam and TomTom. In January 2026, I publicly defended my thesis titled Visual Understanding of Dynamic Scenes using Object Relationships and Open Vocabularies. I was supervised by Prof. dr. Theo Gevers, dr. Martin Oswald and dr. Sezer Karaoglu. Next to my research, I worked as a teaching assistant in the courses Computer Vision 1, Computer Vision 2 and have supervised master AI students Maksymilian Kulicki and Robin Sasse.

My PhD research mainly focused on open-ended perception in 2D, 3D and video without a human-in-the-loop. Our methods automatically generate target vocabularies from the data itself, allowing for scene-specific, flexible vocabulary adaptation without requiring manual prompting of the vocabulary. We coin these methods Auto-Vocabulary or Vocabulary-Free. A second line of work focused on leveraging relational properties for downstream tasks such as instance segmentation.

Currently, I work as a Machine Learning/Computer Vision Engineer in the biotechnology sector to accelerate drug discovery for neurodegenerative diseases with computer vision.

Simultenously, I am building open-ended perception models for real-world use with my own start-up, Unseen AI.

Publications

Auto-Vocabulary Semantic Segmentation. Ulger, O., Kulicki, M., Asano, Y., Oswald, M. R., 2023. In ICCV2025 [paper] [code]

3D-AVS: Auto-Vocabulary Segmentation for LiDAR Points. Wei, W.*, Ulger, O.*, Najadasl, F. K., Gevers, T., Oswald, M. W., 2025. In CVPR2025 [paper] [code]

Relational Prior Knowledge Graphs for Detection and Instance Segmentation. Ulger, O., Wang, Y., Galama, Y., Karaoglu, S., Gevers, T., Oswald, M. R., 2023. In ICCV2023 SG2RL [paper] [code]

Multi-Task Edge Prediction in Temporally-Dynamic Video Graphs. Ulger, O., Wiederer, J., Ghafoorian, M., Belagiannis, V., Mettes, P., 2022. In BMVC2022 [paper]

Evaluating the Effectiveness of Adversarial Patch Attacks on Convolutional Neural Networks. Ulger, O., Stol, M., Mensink, T., 2018. In BNAIC2018 [paper]

*equal contribution.

In Other News..

2025 — Outstanding Reviewer, CVPR | Hawaï, USA
I was selected as an outstanding reviewer.

2024 — Invited Speaker, Nissan Motor Corporation | San Francisco, USA
I gave a talk on open-ended perception and vocabulary-free segmentation.