Obtaining Point cloud range for FRONT_CAM

Hi, I’ve converted the nuscenes-dataset to KITTI-format using “export_kitti.py”.
The question I have is how do I obtain the point cloud range for the points visible in the front camera. To elaborate; a framework I’m using is built on the kitti-format, and they’ve picked the span of 0<x<70.4m, -40<y<40m and -3<z<1m in LiDAR coordinate for their detection of cars. Of course these spans are only optimal for the kitti-dataset. How can i obtain similar data with the converted nuscenes data?
Thanks! :slight_smile:

Hi. Why don’t you just hack a few lines into the export_kitti.py to keep only data within that range?

I’m using a sensor fusion approach with the front camera and I don’t want to exclude points that are in the viability of the camera. Do you know of a simple way to exclude all points but the ones in the visibility of front camera, or how to get the “span” of the lidar points in the front camera view?
Would greatly appreciate it.

Take a look at the logic in https://github.com/nutonomy/nuscenes-devkit/blob/master/python-sdk/nuscenes/utils/geometry_utils.py#L67. You need to project it to the camera, see if it falls within the image and also check that it’s in front.