This project aimed to create a sophisticated environment perception system for autonomous vehicles. By integrating advanced sensor fusion, object detection, tracking, and semantic/instance segmentation algorithms, we will develop high-definition environment maps crucial for safe and efficient navigation. Our focus was on achieving real-time performance, high accuracy, and robustness in challenging driving conditions. Key components include:

Sensor Fusion: Combining data from various sensors (LiDAR, radar, cameras) to create a unified and comprehensive perception of the environment.

Object Detection: Accurately identifying and classifying static and dynamic objects within the scene, such as vehicles, pedestrians, cyclists, and road infrastructure.

Object Tracking: Maintaining continuous tracking of detected objects, predicting their trajectories, and estimating future states.

Semantic/Instance Segmentation: Assigning semantic labels to different parts of the environment and distinguishing individual instances of objects.

High-Definition Map Building: Generating detailed and up-to-date maps of the environment, including road geometry, lane markings, traffic signs, and obstacles.