ApolloCar3D

ApolloCar3D是由百度研究院创建的大型3D汽车实例理解基准数据集，专为自动驾驶研究设计。该数据集包含5,277张驾驶图像和超过60,000个汽车实例，每个汽车实例都配备了工业级的3D CAD模型，并带有绝对模型尺寸和语义标记的关键点。数据集的规模远超现有的PASCAL3D+和KITTI数据集。创建过程中，考虑了2D-3D关键点对应关系和多个实例间的3D关系，以实现高效的标注。该数据集主要用于自动驾驶领域的3D汽车理解和姿态估计研究，旨在解决自动驾驶中的3D物体识别和姿态估计问题。

ApolloCar3D is a large-scale 3D automotive instance understanding benchmark dataset developed by Baidu Research, tailored specifically for autonomous driving research. Comprising 5,277 driving images and over 60,000 car instances, each instance in this dataset is paired with industrial-grade 3D CAD models, absolute model dimensions, and semantically annotated key points. The scale of ApolloCar3D significantly outperforms existing datasets including PASCAL3D+ and KITTI. During the dataset construction, 2D-3D key point correspondence and 3D spatial relationships between multiple instances were taken into account to facilitate efficient annotation. This dataset is primarily dedicated to 3D automotive understanding and pose estimation research in the autonomous driving domain, with the goal of solving the challenges of 3D object recognition and pose estimation in autonomous driving scenarios.