Where do the values in arch_cfg.yaml file come from, especially fov_up, fov_down, img_means, img_stds?

[finnSartoris]

Where do the values fov_up=3, fov_down=-25 and img_means and img_stds in arch_cfg.yaml file come from? Since I currently have a model applied to the SemanticKitti dataset that uses these values and now want to apply it to another dataset, I am trying to figure out if I can reuse these values. Are these values only dependent on the scanner that is used? So if a scanner of the same type (here: Velodyne HDL64) was used, are the above values also the same or what do these values depend on? Thanks a lot!

[jbehley]

The fov_up and fov_down are sensor-specific and are the field of view of the Velodyne HDL-64E. The img_mean and img_stds are computed from the range images. As noted in https://github.com/PRBonn/lidar-bonnetal/issues/43, these values are computed on the training data. However, it's somehow lost how these values are exactly are computed.

And I guess this question is more related to lidar-bonnetal?

[finnSartoris]

Thanks for the quick reply, helps me a lot! Yes that's right, this issue is more related to lidar-bonnetal. Does signal mean the remission of a point or what does the value for signal in img_means and img_std stand for?

[jbehley]

signal is the remission of the sensor.

[finnSartoris]

Great. Thanks a lot!