nuScenes-ODD: An Extension of the nuScenes Dataset with Operational Design Domain Annotations

nuScenes-ODD is an extension of the widely used nuScenes dataset, providing additional annotations for Operational Design Domain (ODD) parameters relevant to autonomous driving research. While nuScenes offers multimodal sensor data (camera, LiDAR, radar) and 3D object annotations, it does not explicitly encode contextual driving conditions. This extension enriches nuScenes with structured ODD attributes, enabling research on scenario-based testing, validation, and decision-making under varying operational contexts. The dataset is provided in tabular CSV format. Each row is indexed by the sample_token, linking directly to nuScenes keyframes (2 Hz). Scene-level attributes that remain constant over an entire sequence are cross-referenced using the scene_token. The extended annotations include: Scenery: road type, lane configuration, drivable area, regulatory/warning/information traffic signs with time-of-operation... Environmental Conditions: weather (rain, fog), visibility, lighting... Dynamic Elements: traffic density, unusual road users, temporal smoothing of categorical attributes...nuScenes-ODD是当前广泛使用的nuScenes数据集的扩展版本，为自动驾驶研究相关的运行设计域（Operational Design Domain，ODD）参数提供了额外标注。原nuScenes数据集虽已提供多模态传感器数据（摄像头、激光雷达、雷达）及3D目标标注，但未对上下文驾驶场景信息进行显式编码。 该扩展数据集为nuScenes补充了结构化的ODD属性，支持在多样化运行场景下开展基于场景的测试、验证与决策相关研究。 本数据集以表格型CSV格式提供，每一行通过样本令牌（sample_token）进行索引，可直接关联至nuScenes的关键帧（帧率为2Hz）。对于在整个序列中保持恒定的场景级属性，则通过场景令牌（scene_token）进行交叉引用。 扩展标注涵盖以下类别： - 场景要素：道路类型、车道配置、可行驶区域、带有运营时段的管制/警示/信息类交通标志…… - 环境条件：天气（降雨、雾霾）、能见度、光照情况…… - 动态要素：交通密度、异常道路参与者、分类属性的时域平滑处理……