CenterPoint model trained on livox dataset and deployed with TensorRT on ros2. Code is written according to this project.
Overall inference has three phases:
the project is running inference on LivoxOpenDataSet.
The .pt file you can get in this project. the onnx file is provided by this project. You can also get the onnx file by yourself through programming, but you have to consider the problem that onnx does not support atan2.
TensorRT and cuda are necessary conditions for running centerpoint
| Function(unit:ms) | NVIDIA RTX 3060 Laptop GPU | NVIDIA Jetson AGX Orin |
| ----------------- | --------------------------- | --------------------------- |
| Preprocess | 0.044534 ms | 0.323782 ms |
| Rpn | 25.0291 ms | 48.5475 ms |
| Postprocess | 0.80967 ms | 1.8511 ms |
| Summary | 25.8914 ms | 50.7353 ms |