DeepLoc, Oxford Robot-Car, APR-BeIntelli
收藏资源简介:
本文介绍了一个名为APR-Transformer的模型架构,它使用图像或激光雷达数据预测绝对姿态(3D位置和3D方向)。论文中提及了三个数据集:DeepLoc、Oxford Robot-Car和APR-BeIntelli。DeepLoc是一个大规模的城市户外定位数据集,包含RGB图像和通过激光雷达SLAM系统计算的精确姿态标签。Oxford Robot-Car是一个包含激光雷达、相机、GNSS和雷达信息的多模态户外数据集,用于评估模型在各种天气、交通和行人条件下的性能。APR-BeIntelli是一个自定义数据集,由柏林工业大学DAI实验室在BeIntelli项目中收集,包含多摄像头RGB图像和同步的激光雷达点云,用于测试APR-Transformer在动态和复杂场景中的实用性。
This paper introduces a model architecture named APR-Transformer, which predicts absolute pose (3D position and 3D orientation) using image or LiDAR data. Three datasets are mentioned in this paper: DeepLoc, Oxford Robot-Car, and APR-BeIntelli. DeepLoc is a large-scale urban outdoor localization dataset containing RGB images and accurate pose labels calculated via LiDAR SLAM systems. The Oxford Robot-Car is a multimodal outdoor dataset that integrates LiDAR, camera, GNSS and radar information, and is used to evaluate model performance under various weather, traffic and pedestrian conditions. APR-BeIntelli is a custom dataset collected by the DAI Lab of Technische Universität Berlin under the BeIntelli project, which includes multi-camera RGB images and synchronized LiDAR point clouds for testing the practicality of APR-Transformer in dynamic and complex scenarios.




