CARLA, nuScenes, UGV
收藏资源简介:
本文构建了三种数据集:CARLA数据集,通过CARLA模拟器生成,包含多个配备LiDAR和相机传感器的模拟代理;nuScenes数据集,包含大规模的LiDAR扫描和对象注释;UGV数据集,通过实际部署的无人地面车辆收集,具有静态和动态遮挡的挑战性条件。这些数据集支持视场估计的研究,并为评估FOV估计算法在对抗环境下的鲁棒性提供了基准。
In this paper, three datasets are constructed: 1. The CARLA dataset: generated via the CARLA simulator, which contains multiple simulated agents equipped with LiDAR and camera sensors; 2. The nuScenes dataset: comprising large-scale LiDAR scans and object annotations; 3. The UGV dataset: collected using a field-deployed unmanned ground vehicle, with challenging scenarios involving static and dynamic occlusions. These datasets support research on field of view (FOV) estimation and provide benchmarks for evaluating the robustness of FOV estimation algorithms under adversarial environments.

- 1Probabilistic Segmentation for Robust Field of View Estimation杜克大学电气与计算机工程系 · 2025年



