TrafficScene
收藏资源简介:
TrafficScene 数据集是首个将光场图像和激光雷达点云数据结合的语义分割数据集。它使用一个具有 30 厘米基线的 3x3 FLIR BFS-PGE 16S2C 相机阵列和 CH128X1 激光雷达在交通场景中收集数据,包含了 5607 张光场图像和 623 帧 点云数据。每个点云帧包含超过 60,000 个点。所有光场视图都被校正并使用 CVAT 工具进行了 15 个语义类别的标注,这是首个在两个模态上都有对齐语义标签的多模态数据集。TrafficScene 数据集涵盖了五个典型的交通场景和五种常见的交通参与者,对于自动驾驶和智能交通系统具有重大价值。
The TrafficScene dataset is the first semantic segmentation dataset that combines light field images and LiDAR point cloud data. Data was collected in traffic scenarios using a 3x3 FLIR BFS-PGE 16S2C camera array with a 30 cm baseline and a CH128X1 LiDAR. The dataset contains 5,607 light field images and 623 point cloud frames. Each point cloud frame contains over 60,000 points. All light field views have been calibrated and annotated with 15 semantic categories using the CVAT tool. This is the first multimodal dataset with aligned semantic labels across both modalities. The TrafficScene dataset covers five typical traffic scenarios and five common traffic participants, which is of great value to autonomous driving and intelligent transportation systems.

- 1Semantic Segmentation Algorithm Based on Light Field and LiDAR Fusion清华大学深圳国际研究生院 · 2025年



