Differenes between radar point cloud data and processed data


  1. I use the function ‘RadarPointCloud.from_file(data_path)’ to get radar point cloud with a size(18,125)
  2. Then I use ‘nusc.get_sample_data(radar_data[‘token’])’ and get the processed data.

I have thought only some filter was made, but the specific data are completely different, e.g. ‘z’ value, how to understand the difference, and why the latter doesn’t contain velocity information.

Thanks a lot.

Take a look at the code:

“Note that the boxes are transformed into the current sensor’s coordinate frame.”
The velocity isn’t used here. You can however look at the vx/vy values in https://github.com/nutonomy/nuscenes-devkit/blob/master/python-sdk/nuscenes/utils/data_classes.py#L292

Thank you for your reply.

By the way, I still have a question, is it possible to make sure the relationship bewteen each radar point and actual object ?

  1. With the function “get_sampe_data”, it’s easy for each point to find the annotation box with tokens, but it lacked velocity information.

  2. To get the velocity information, I have to fetch the point data directly from files. Is there any one-to-one match between these points with boxes?

The easiest would probably be to create a copy of get_sample_data that does not throw away the velocity information.