一些疑惑

[cjl-2000]

您工作中的无人机位姿估计方法让我收益匪浅，我有几个问题希望得到您的解答：
1.请问该工作能否扩展至其他未知场景？不同的场景下是否必须要通过重新训练才能正常运行？因为基于3D纹理地图的定位方法往往不需要进行微调即可获得较好的效果。
2.LoD地图的获取是否存在阻碍？
3.算法的推理速度？

[VictorZoo]

Hi @cjl-2000 ,

1. 对于扩展至未知场景,请参考我们的补充材料的表格13, https://victorzoo.github.io/LoD-Loc.github.io/files/LoD_Loc_Suply.pdf ;
2. 市面上有很多公开的LoD地图,并提供下载链接.我们正在整理中,敬请期待;
3. 算法的推理速度请参考补充材料的表格14.

Victor

---

Hi @cjl-2000 ,

1. For experiments on generalization to unknown scenarios, please refer to Table 13 in our supplementary materials (https://victorzoo.github.io/LoD-Loc.github.io/files/LoD_Loc_Suply.pdf).
2. There are many publicly available LoD maps with download links. We are currently organizing these resources—stay tuned!
3. For the inference speed of the algorithm, please refer to Table 14 in the supplementary materials.

Victor