search

xiaoaoran / FPS-Net

Code for "FPS-Net: A convolutional fusion network for large-scale LiDAR point cloud segmentation".
MIT License
21 stars 6 forks source link

A question about depth and xyz #4

Closed yuyunlong2002 closed 2 years ago

yuyunlong2002 commented 2 years ago

Hello, I want to ask why we should use depth and xyz at the same time. I observed in your code that the depth is calculated using xyz（depth=np.linalg.norm(self.points, 2, axis=1)）, so there is a correlation between the two. Why do we use depth when using xyz , is there any result indicating that using depth and xyz simultaneously will have better effect?

xiaoaoran commented 2 years ago

Hi @yuyunlong2002

Yes, although depth can be directly calculated from xyz, adding this modality can still enrich input since the network does not implicitly learn depth information. This is the trick also used by existing works such as SqueezeSegV2 and rangenet++.

You can find relevant discussion and analysis in Table 4 of our paper FPS-Net.

xiaoaoran commented 2 years ago

@yuyunlong2002

Hi, I closed this issue since there are no further questions. Please feel free to re-open it if you have any other questions.