Do you have any tests on the performance of the proposed algorithm on large scale autonomous driving scene?

[jiangchaokang]

The experiments in the paper are experiments on indoor or small objects, have you tested its effect on large scale datasets? For example, Nuscnen.

[ZrrSkywalker]

Thanks for your valuable suggestion. We haven't test our approach for outdoor 3D scenes. We will do some experiments and update in the repo.

[jiangchaokang]

Thank you very much for your prompt reply~ I am a master student in Shanghai Jiao Tong University. I am very interested in following your work on inter-frame matching tasks, such as LiDAR odometry, point cloud registration or 3D scene flow. Could you please provide some hints about completing these tasks?

[ZrrSkywalker]

Thanks for your interest. Given the distribution difference, I suppose the local operators or multi-scale hierarchy can be modified to better fit the sparser scene-level point clouds. I'm happy to have further discussion.

[jiangchaokang]

Okay, thank you for your valuable comments!