Leandro Di Bella

I'm a PhD Researcher in Computer Vision and Deep Learning at Vrije Universiteit Brussel (VUB) currently working on monocular 3D object detection and tracking for autonomous driving applications. My research focuses on developing novel algorithms that combine classical filtering techniques with deep learning to improve the temporal consistency and accuracy of vehicle pose estimation and tracking in traffic scenarios.

My recent work includes HybridTrack, a novel approach that integrates data-driven Kalman Filtering within a tracking-by-detection framework, and LAM3D, a transformer-based architecture for monocular 3D object detection. I'm particularly interested in leveraging attention mechanisms and hybrid architectures that combine the best of traditional computer vision techniques with modern deep learning approaches.

A key focus of my research is developing solutions that are not just theoretically sound but also practically applicable in real-world autonomous driving scenarios. This includes ensuring real-time performance while maintaining high accuracy, as demonstrated in my work on DeepKalPose and MonoKalman, which address the critical challenge of temporal consistency in vehicle pose estimation from monocular video sequences.

Publications

HybridTrack: A Hybrid Approach for Robust Multi-Object Tracking

HybridTrack: A Hybrid Approach for Robust Multi-Object Tracking

Leandro Di Bella, Yangxintong Lyu, Bruno Cornelis, Adrian Munteanu

LAM3D: Leveraging Attention for Monocular 3D Object Detection

LAM3D: Leveraging Attention for Monocular 3D Object Detection

Diana-Alexandra Sas, Leandro Di Bella, Yangxintong Lyu, F. Oniga, Adrian Munteanu

IEEE International Workshop on Multimedia Signal Processing 2024

MonoKalman: Monocular Vehicle Pose Estimation with Kalman Filter-based temporal consistency

MonoKalman: Monocular Vehicle Pose Estimation with Kalman Filter-based temporal consistency

Leandro Di Bella, Yangxintong Lyu, Bruno Cornelis, Adrian Munteanu

International Conference on Mobile Data Management 2024

DeepKalPose: An Enhanced Deep-Learning Kalman Filter for Temporally Consistent Monocular Vehicle Pose Estimation

DeepKalPose: An Enhanced Deep-Learning Kalman Filter for Temporally Consistent Monocular Vehicle Pose Estimation

Leandro Di Bella, Yangxintong Lyu, Adrian Munteanu

Electronics Letters 2024