使用TensorRT部署BEVDet-ROS模型C++源码.zip

# BEVDet by TensorRT, C++, ROS <p align="center"> <img src="./doc/1.gif" width="375" height="338" /> </p> + This repository contains source code and models for BEVDet online real-time inference using CUDA, TensorRT, ROS1 & C++. + Modify the record reference [modify](https://github.com/linClubs/BEVDet-ROS-TensorRT/tree/main/doc/modify.md) # 1 ENV - **ubuntu-20.04、CUDA-11.3、cuDNN-8.6.0、TensorRT-8.5** - **yaml-cpp、Eigen3、libjpeg** # 2 Build ~~~python mkdir -p bev_ws/src cd bev_ws/src git clone https://github.com/linClubs/BEVDet-ROS-TensorRT.git cd .. catkin_make source devel/setup.bash ~~~ # 3 Run 1. generate engine Generate the onnx TensorRT engine reference to [bevdet-tensorrt-cpp](https://github.com/LCH1238/bevdet-tensorrt-cpp) engine path as follows. ~~~python # engine path BEVDet-ROS-TensorRT └──ckpts ├── bev_stage_lt_d.engine ├── img_stage_lt_d.engine └── lt_d.yaml ~~~ 2. data preparation The rosbag folder can be downloaded from [Baidu Netdisk](https://pan.baidu.com/s/1f3nUnHa_4cd6FsRTV8YhkA?pwd=rjim) 3. demo test ~~~python # 1. start bevdet_node roslaunch bevdet bevdet_node.launch # 2 play data rosbag play nus.bag ~~~ <p align="center"> <img src="./doc/1.png" width="800" height="720" /> </p> --- # References - [bevdet-tensorrt-cpp](https://github.com/LCH1238/bevdet-tensorrt-cpp) - [BEVDet](https://github.com/HuangJunJie2017/BEVDet) - [mmdetection3d](https://github.com/open-mmlab/mmdetection3d) - [nuScenes](https://www.nuscenes.org/) - bev感知交流群-472648720, 欢迎各位小伙伴进群一起学习讨论bev相关知识！！！^_^ <p align="center"> <img src="./doc/1.jpg" width="200" height="200" /> </p> ---