SemanticKITTI-C, SemanticPOSS-C
收藏arXiv2024-01-07 更新2024-08-06 收录
下载链接：
http://arxiv.org/abs/2301.00970v3
下载链接
链接失效反馈官方服务：
资源简介：
SemanticKITTI-C和SemanticPOSS-C是专为评估激光雷达语义分割模型在各种损坏下的鲁棒性而设计的大型基准数据集。这些数据集包含4071和500个激光雷达扫描，分别对应于SemanticKITTI和SemanticPOSS的验证集，并引入了16种不同类型的损坏，包括不利天气条件、测量噪声和跨设备差异。这些损坏旨在模拟自动驾驶等安全关键应用中可能遇到的真实世界情况，从而全面分析和提高激光雷达语义分割模型的鲁棒性和泛化能力。
SemanticKITTI-C and SemanticPOSS-C are large-scale benchmark datasets specifically designed to evaluate the robustness of LiDAR semantic segmentation models against various corruptions. These datasets comprise 4071 and 500 LiDAR scans, corresponding respectively to the validation sets of SemanticKITTI and SemanticPOSS. They incorporate 16 distinct types of corruptions, including adverse weather conditions, measurement noise, and cross-device discrepancies. These corruptions are intended to simulate real-world scenarios that may be encountered in safety-critical applications such as autonomous driving, enabling comprehensive analysis and enhancement of the robustness and generalization capabilities of LiDAR semantic segmentation models.
提供机构：
香港中文大学（深圳）
创建时间：
2023-01-03
