Skip to content

Latest commit

 

History

History
67 lines (42 loc) · 2.19 KB

README.md

File metadata and controls

67 lines (42 loc) · 2.19 KB

PyTorch Implementation of Center Based Lidar 3D Object Detection for Autonomous Driving

Solutions Architect: Gaowei Xu (gaowexu1991@gmail.com)

1. Introduction

The Lidar sensor setup could be referred as the illustration figure below:

lidar_coord_origin

The ego front is x-axis, the left side is y-axis and the upwards is z-axis. Each position of environment agent is encoded with a 7-dimension vector, i.e., (x, y, z, length, width, height, orientation). orientation is in range (-pi, pi], which is illustrated in above figure.

CenterPoint use point pillar to extract points cloud's features and then scatter the features to pseudo image, the figure below illustrate the global indices detail between real physical world and pseudo image space. The global indexing method used in scatter module should be the same with the ground truth (heatmap) generation logic.

voxelization_and_global_index

2. Setup Environment

Install Nvidia Driver and CUDA 11.6

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin
sudo mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600
wget https://developer.download.nvidia.com/compute/cuda/11.3.0/local_installers/cuda-repo-ubuntu2004-11-3-local_11.3.0-465.19.01-1_amd64.deb
sudo dpkg -i cuda-repo-ubuntu2004-11-3-local_11.3.0-465.19.01-1_amd64.deb
sudo apt-key add /var/cuda-repo-ubuntu2004-11-3-local/7fa2af80.pub
sudo apt-get update
sudo apt-get -y install cuda

Install System Dependencies

sudo apt-get update
sudo apt-get install -y libopencv-dev python3-pip git

Install Python Packages

Download torch-1.10.2+cu113-cp38-cp38-linux_x86_64.whl to local disk and then install it:

pip3 install torch-1.10.2+cu113-cp38-cp38-linux_x86_64.whl
pip3 install -r requirements.txt

3. Model Training

4. Performance Evaluation

5. ONNX Conversion

6. TensorRT Conversion

7. CUDA C++ Integration

8. License

See the LICENSE file for our project's licensing. We will ask you to confirm the licensing of your contribution.