Hi, I’m Mengyu. I hold an M.Sc. degree from the Technical University of Munich. My academic interests lie in event-based vision for scene understanding, particularly in the context of autonomous driving and intelligent transportation systems. I am currently seeking a PhD opportunity where I can further explore the potential of event-based vision and contribute to cutting-edge research in this exciting field.

Master’s Thesis

Multi-Modal Fusion of Image Sequences for Dense Prediction with RGB and Event Cameras in Autonomous Driving

My Master’s thesis explores the integration of RGB and event camera data to enhance dense prediction tasks—such as semantic segmentation, depth estimation, and object detection—in autonomous driving. RGB cameras deliver high-resolution visual information essential for scene understanding, while event cameras provide high temporal resolution and dynamic range, enabling motion detection and perception under extreme lighting conditions. By fusing these complementary modalities, the system can construct a more complete and resilient representation of dynamic traffic environments. This research investigates multi-modal feature fusion strategies tailored for semantic segmentation.

Semester Thesis

Event-Based Multi-Object Tracking Dataset and Baseline for Traffic Scenarios

Conventional multi-object tracking in Intelligent Transportation Systems (ITS) largely relies on frame-based cameras, which often suffer from poor performance under challenging conditions such as low illumination and high-speed motion. To address these limitations, my research focuses on leveraging event cameras—sensors characterized by low latency, high dynamic range, and high temporal resolution—to enable robust tracking in complex traffic environments. As part of this effort, I developed the TUMTraf EMOT dataset, the first event-based dataset tailored to ITS scenarios, encompassing separate tracking sequences for vehicles and pedestrians. Alongside the dataset, I proposed a tracking-by-detection baseline with a specialized feature extractor and pre-trained models, balancing efficiency and accuracy. Experimental results demonstrate the strong performance of our method, establishing a reliable benchmark that encourages further exploration of event-based perception in ITS.

Internship

Decentralized Tracking in the Context of the DISRUPT Project

As part of the DISRUPT project, my internship focused on decentralized multi-object tracking for intelligent transportation systems. I investigated multi-camera object tracking using a large-scale synthetic dataset, comparing decentralized and centralized tracking algorithms. I further explored covariance intersection methods to enhance decentralized fusion performance and tracking reliability across distributed sensor nodes. I also generated ground truth data from vehicle test drives using GNSS and RTK (real-time kinematics) technologies, and actively participated in field testing for system validation.

From a systems integration perspective, I designed and implemented an MQTT-based communication pipeline to enable real-time data transmission between heterogeneous edge devices and the cloud. Additionally, I conceptualized efficient data exchange protocols in a ROS2 wireless network, and created Dockerized modules to ensure scalable and hardware-independent deployment of sensor fusion components.