TartanAir
收藏资源简介:
该数据集名为TartanAir,被用于训练,并且是2020年ECCV视觉同步定位与地图构建(SLAM)竞赛的一部分。它包含了一个测试集,用于评估AutoLoop模型与现有最先进方法的性能。在TartanAir上进行的离线预计算生成了551个闭环匹配对,实验是在包含闭环匹配对和非闭环帧的序列上进行的。该数据集包含了多个对SLAM具有挑战性的序列,其任务包括视觉SLAM评估和闭环检测。
This dataset, named TartanAir, is utilized for training and was part of the 2020 ECCV Visual Simultaneous Localization and Mapping (SLAM) Competition. It features a test set for assessing the performance of the AutoLoop model against state-of-the-art baseline methods. Offline precomputation on the TartanAir dataset yields 551 loop closure matching pairs, with experiments conducted on sequences that incorporate both these loop closure pairs and non-loop closure frames. This dataset includes multiple SLAM-challenging sequences, with tasks covering visual SLAM evaluation and loop closure detection.

- 1TartanAir: A Dataset to Push the Limits of Visual SLAM卡内基梅隆大学机器人研究所 · 2020年



