How to preprocess pointcloud for PointPillars Inference in C++ TensorRT

[Acuno41]

Greetings,

I successfully generated tensorRT .engine model from sample PointPillars model from mmdetection3d and tested it with tools/test.py program.
Then I started writing inference code using tensorRT in c++ and I successfully loaded and deserialized generated .engine model with mmdeploy custom_ops.

My question is, how should I preprocess or rearrange pointcloud data in order to feed it to the model ? Is there any example of how you feed pointcloud data to lidar based object detection methods ?