Mohammadjavad completed his Ph.D. in the lab in September 2026. His dissertation on route-structured perception for railway scene understanding was recommended for a Dissertation Prize. As a postdoctoral fellow, he works on AI for autonomous trains and railway perception.
Current projects
- Ego-path segmentation for autonomous trains with vision-language models (ISPRS 2026)
- Sidewalk delivery robots in a campus digital twin
Publications (7)
2026
- Ghorbanalivakili, M., Varghese, A., and Sohn, G. (2026). Reasoning-guided Ego-path Segmentation for Autonomous Trains using Vision-Language Models. Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLIX-B3-2026, pp. 89-96. doi:10.5194/isprs-archives-XLIX-B3-2026-89-2026
- Varghese, A., Ghorbanalivakili, M., and Sohn, G. (2026). The Emerging Role of Vision-Language Models in the Automation of Railway Asset Management: A Review and Future Perspective. Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLIX-B2-2026. doi:10.5194/isprs-archives-XLIX-B2-2026-745-2026
2025
- Jung, J., Ghorbanalivakili, M., and Sohn, G. (2025). Multi-Railway Track and Switch Region Recognition Using Mobile Laser Scanning Data. IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), 3713-3720.
- Ghorbanalivakili, M. and Sohn, G. (2025). TRIT-Net: Triplet-based Railway Instance Tracing Network Using Attraction Field Representation. Proceedings of the Conference on Robots and Vision.
2023
- Ghorbanalivakili, M., Kang, J., Sohn, G., Beach, D., and Marin, V. (2023). TPE-Net: Track Point Extraction and Association Network for Rail Path Proposal Generation. IEEE 19th International Conference on Automation Science and Engineering (CASE).
Preprints and Under Review
- Sohn, G., Ghorbanalivakili, M., and Varghese, A. Triplet-based Railway Ego-path Instance Tracing Using Learned Attraction Field Representation. SSRN 6827662.
- Alizadeh Naeini, A., Sheikholeslami, M. M., Ghorbanalivakili, M., Yoo, S., and Sohn, G. A Unified Multi-Task Learning Framework for DTM Generation Using Surface Differencing. SSRN 5045901.