CV

Summary

PhD candidate at ETRO (VUB) and imec, specialized in robotics, physical AI, and computer vision for autonomous systems. Currently technical lead on drone localization and navigation in a European Defence Fund (EDF) research project (ORIGAMI), and working on multi-agent motion forecasting using flow-based generative AI. Experienced in deploying real-time perception on embedded platforms (Jetson, TensorRT, C++).

Research interests

Perception for autonomous systems, world models for perception and navigation, localization in 3D environments, vision-language-action (VLA) models for robotics.

Personal data

• Location: Brussels, Belgium
• Email: leandro.dibella@gmail.com
• GitHub: https://github.com/leandro-svg
• Google Scholar: https://scholar.google.de/citations?user=f7IDHsgAAAAJ&hl=en
• ORCID: https://orcid.org/0009-0000-1731-7205
• LinkedIn: https://www.linkedin.com/in/leandro-di-bella-62381413b/
• App: https://mappx.app

Education

• 2023–Present: PhD in Engineering Sciences (AI, Computer Vision & Robotics) — Department of Electronics and Informatics (ETRO), Vrije Universiteit Brussel (VUB)
• 2026: ICVSS International Computer Vision Summer School — “Computer Vision for Spatial and Physical Intelligence” (Sicily, Italy)
• 2025–Present: Advanced Master in Industrial and Technological Management — Solvay Brussels School
• 2020–2023: Master Electrical Engineer (Information Technology Systems option) — Bruface Faculty of Engineering (ULB/VUB)
• 2017–2020: Bachelor Engineer — Bruface Faculty of Engineering (ULB/VUB)

Work experience

• Technical Lead, Robotics & Perception Researcher — IMEC & VUB, ORIGAMI (European Defence Fund project) (Jan 2026–Present)
  • Technical lead for drone autonomy: architected and integrated the full drone platform (hardware, sensing, and autonomy software)
  • Sensor fusion and localization in 3D environments; developing world-model-based methods for autonomous navigation
• Computer Vision Researcher — MACQ Mobility (research project) (Jan 2023–Jan 2026)
  • Detection, 3D multi-object tracking, scene understanding, and motion forecasting for autonomous vehicles using generative AI
  • Achieved #1 on the KITTI MOT and Waymo motion forecasting benchmarks
• Founder — Mappx (Aug 2025–Present)
  • Launched Mappx mobile app in Aug 2025
  • Developed full-stack Flutter frontend and FastAPI backend, integrating location-based social sharing with maps and photos
  • Implemented cost-effective backend infrastructure using Azure and Firebase to support scalable user engagement and deployment on the App Store
• Internship — MACQ Mobility (Aug 2022–Oct 2022)
  • Developed and integrated instance segmentation on a Jetson TX2 edge device (Python/C++)

Projects

• 2023: Embedded AI — Real-Time Instance Segmentation with TensorRT and ONNX Deployment
  • Implemented and optimized SparseInst and Yolov7 for real-time deployment on NVIDIA edge devices using CUDA TensorRT

Languages

• French (Native)
• English (C1)
• Dutch (B1)

Skills

• Programming: Python, C++, CUDA, PyTorch, TensorRT, ONNX, FastAPI, Flutter
• Robotics: ROS 2, SLAM, localization, motion planning, navigation, sensor fusion
• ML / Perception: detection, instance segmentation, monocular 3D, 2D/3D multi-object tracking, Kalman filtering, temporal consistency
• Vision-language & GenAI: grounding / referring tracking, multimodal reasoning, generative motion forecasting
• Tools/Platforms: Jetson, Docker, Azure Cloud Services, Firebase, GitHub CI/CD
• Soft skills: Team spirit, ownership/responsibility, comfort zone growth, adaptability

Publications

Books

Journal Articles

• FlowS: One-Step Motion Prediction via Local Transport Conditioning
  Di Bella, L., Munteanu, A., & Cornelis, B. (2026). “FlowS: One-Step Motion Prediction via Local Transport Conditioning.” IEEE Robotics and Automation Letters (RA-L), accepted.
• HybridTrack: A Hybrid Approach for Robust Multi-Object Tracking
  Di Bella, L., Lyu, Y., Cornelis, B., & Munteanu, A. (2025). “HybridTrack: A Hybrid Approach for Robust Multi-Object Tracking.” IEEE Robotics and Automation Letters (RA-L/ICRA).
• DeepKalPose: An Enhanced Deep-Learning Kalman Filter for Temporally Consistent Monocular Vehicle Pose Estimation
  Di Bella, L., Lyu, Y., & Munteanu, A. (2024). “DeepKalPose: An Enhanced Deep-Learning Kalman Filter for Temporally Consistent Monocular Vehicle Pose Estimation.” Electronics Letters.
• Automating Coral Reef Fish Family Identification on Video Transects Using a YOLOv8-Based Deep Learning Pipeline
  Gerard, J., Di Bella, L., Huyghe, F., & Kochzius, M. (2025). “Automating Coral Reef Fish Family Identification on Video Transects Using a YOLOv8-Based Deep Learning Pipeline.” arXiv preprint, arXiv:2511.00022.

Conference Papers

• Spectral-Aware Multi-Object Tracking in Harsh Aerial Perception Domains
  Di Bella, L., Mimassi, J., Denis, L., & Munteanu, A. (2026). “Spectral-Aware Multi-Object Tracking in Harsh Aerial Perception Domains.” IEEE ICRA 2026 Workshop S2S.
• ChronoFusion: Spatio-Temporal Super-Resolution based on Graph VAEs and Gated Fusion
  Moghadas, S. M., Di Bella, L., Cornelis, B., & Munteanu, A. (2025). “ChronoFusion: Spatio-Temporal Super-Resolution based on Graph VAEs and Gated Fusion.” European Workshop on Visual Information Processing (EUVIP).
• ReferGPT: Towards Zero-Shot Referring Multi-Object Tracking
  Chamiti, T.*, Di Bella, L.*, Munteanu, A., & Deligiannis, N. (2025). “ReferGPT: Towards Zero-Shot Referring Multi-Object Tracking.” Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW).
• LAM3D: Leveraging Attention for Monocular 3D Object Detection
  Sas, D.-A., Di Bella, L., Lyu, Y., Oniga, F., & Munteanu, A. (2024). “LAM3D: Leveraging Attention for Monocular 3D Object Detection.” 2024 IEEE 26th International Workshop on Multimedia Signal Processing (MMSP), pp. 1–6. IEEE.
• Monokalman: Monocular Vehicle Pose Estimation with Kalman Filter-Based Temporal Consistency
  Di Bella, L., Lyu, Y., Cornelis, B., & Munteanu, A. (2024). “Monokalman: Monocular Vehicle Pose Estimation with Kalman Filter-Based Temporal Consistency.” IEEE International Conference on Mobile Data Management (MDM).

Talks

• Merging AI and deterministic approaches for better performance: AI-Enhanced Kalman Filtering for Robust Tracking — January 2024 — AutoSens 2024

Teaching

• Teaching Assistant (2.5 years, 2022–2025) — Machine Learning and Big Data Processing, Vrije Universiteit Brussel (VUB)

Supervision

• Supervised 6 Master’s theses:
  • Jules Gerard — Automating Coral Reef Fish Family Identification on Video Transects Using a YOLOv8-Based Deep Learning Pipeline
  • Diana Alexandra Sas — Monocular 3D Object Detection with Pyramid Vision Transformer
  • Kyan David — Autonomous Drone Navigation in GNSS-Degraded and Non-Permissive Environments
  • Mohammed Marsour — Pruning and Quantization Strategies for Efficient Camera-Based Object Detection
  • Cristian Vladoiu — Pedestrian Intention Prediction via Vision-Language Action Models
  • Mayur Ashok Sonawane — Object Pose Estimation on Embedded Devices