Congratulations to Dr. Mohammadjavad Ghorbanalivakili, who successfully defended his Ph.D. dissertation in Earth and Space Science at York University on September 21, 2026.
His dissertation, “Route-Structured Perception for Railway Scene Understanding Using Forward-Looking Camera Data,” develops camera-based methods that help autonomous trains understand which tracks lie ahead in complex multi-track and switch scenes. It makes three connected contributions:
- TPE-Net, a triplet-based representation that pairs left and right rails into possible path proposals;
- TRIT-Net, which traces each ego-path instance through a learned Attraction Field Map instead of hand-made rules;
- reasoning-guided valid-route segmentation with a vision–language model, which identifies the route the train will take through a switch and can explain its choice.
The work also released RailSem19+, a corrected version of the RailSem19 benchmark with 7,226 repaired switch-path annotations.
The examining committee recommended the dissertation for a Dissertation Prize. We thank the committee members, Prof. Mojgan Jadidi (Chair), Prof. Rongjun Qin, Prof. Jinjun Shan, Prof. Hina Tabassum and Prof. Regina Lee, for their careful reading and thoughtful questions.
Mohammadjavad continues in the lab as a Postdoctoral Fellow, working on AI for autonomous trains and railway perception.
