Inquiry regarding a custom dataset

[sputnik4]

Good morning,

I hope you are all well.

Thank you for your great work on 3D point ground segmentation.

The work you did generalizes well on our custom dataset. However, I have a small challenge, the number of input points of an entire sequence of my dataset is 12 million points (/pointcloud topic sends 12 million points), but the number of output points I get back after inference is 7 million points (/groundgrid/segmented_cloud topic returns 7 million points). What could be the cause of this?

Regards and Thanks.

[Nussknacker]

Hi @sputnik4,

I'm glad to hear that it also works with your data.

Regarding the missing points: Perhaps this is caused by the size preset of the grid map, which is set to 120 m in length and width, with the sensor origin in the center. Points that fall outside of the grid map are dropped. If you want to modify the grid map size, you need to change the value in line 71 of the file include/groundgrid/GroundGrid.h. However, this might impact runtime performance.

[sputnik4]

Hi @Nussknacker

Thank you for your fast response. I am going to give it a try.

Regards.