A project demonstrating how to use CUDA-PointPillars to deal with cloud points data from lidar.
NVIDIA-AI-IOT/CUDA-PointPillars is ranked 15,366 with 659 stars.
Same-category comparison
NVIDIA-AI-IOT/CUDA-PointPillars is ranked 15,366 among these 8 peers; MorvanZhou/pytorch-A3C is ranked 15,365.
Same-category comparison| Name | Rank | Metrics |
|---|
| bbaaii/DreamDiffusion | 15,356 | - Language:
- Python
- Stars:
- 660
- Forks:
- 80
- Description:
- Implementation of “DreamDiffusion: Generating High-Quality Images from Brain EEG Signals”.
|
|---|
| v-iashin/video_features | 15,360 | - Language:
- Python
- Stars:
- 660
- Forks:
- 112
- Description:
- Extract video features from raw videos using multiple GPUs. We support RAFT flow frames as well as S3D, I3D, R(2+1)D, VGGish, CLIP, and TIMM models.
|
|---|
| yzhao062/combo | 15,361 | - Language:
- Python
- Stars:
- 660
- Forks:
- 105
- Description:
- (AAAI' 20) A Python Toolbox for Machine Learning Model Combination.
|
|---|
| MorvanZhou/pytorch-A3C | 15,365 | - Language:
- Python
- Stars:
- 659
- Forks:
- 142
- Description:
- Simple A3C implementation with pytorch + multiprocessing.
|
|---|
| cpnota/autonomous-learning-library | 15,370 | - Language:
- Python
- Stars:
- 659
- Forks:
- 72
- Description:
- A PyTorch library for building deep reinforcement learning agents.
|
|---|
| motiwari/BanditPAM | 15,373 | - Language:
- C++
- Stars:
- 659
- Forks:
- 52
- Description:
- BanditPAM C++ implementation and Python package.
|
|---|
| reservoirpy/reservoirpy | 15,376 | - Language:
- Python
- Stars:
- 659
- Forks:
- 156
- Description:
- A simple and flexible code for Reservoir Computing architectures like Echo State Networks.
|
|---|
| G-U-N/AnimateLCM | 15,378 | - Language:
- Python
- Stars:
- 658
- Forks:
- 46
- Description:
- [SIGGRAPH ASIA 2024 TCS] AnimateLCM: Computation-Efficient Personalized Style Video Generation without Personalized Video Data.
|
|---|
Key parameters and specifications
Key parameters and specifications| Parameter | Value |
|---|
| Language | Python |
|---|
| Category | Model Development |
|---|
| Subcategory | Architecture |
|---|
| Keywords | 3D, Autonomous vehicles, Object detection |
|---|
| Stars | 659 |
|---|
| Forks | 185 |
|---|
| Contributors | 2 |
|---|
| Created date | 2021-11-16 |
|---|
| Updated date | 2026-10-02 |
|---|
| Stars in 1 day | 0 |
|---|
| Stars in 7 days | 0 |
|---|
| Rank | 15,366 |
|---|
| Previous rank | 15,180 |
|---|
| Rank change | -186 |
|---|
| 7-day stars rank | 14,958 |
|---|
| Previous 7-day stars rank | 14,963 |
|---|
| 7-day stars rank change | 5 |
|---|
| Total stars rank | 15,579 |
|---|
| Previous total stars rank | 15,552 |
|---|
| Total stars rank change | -27 |
|---|
| Country | USA |
|---|
README summary
PointPillars Inference with TensorRT This repository contains sources and model for pointpillars inference using TensorRT. Overall inference has below phases: Voxelize points cloud into 10-channel features Run TensorRT engine to get detection feature Parse detection feature and apply NMS Prerequisites Prepare Model && Data We provide a Dockerfile to ease environment setup. Please execute the following command to build the docker image after nvidia-docker installation: We can then run the docker with the following command: For model exporting, please
run the following command to clone pcdet repo and install custom CUDA extensions: Download PTM to ckpts/, then use below command to export ONNX model: Use below command to evaluate on kitti dataset, follow Evaluation on Kitti to get more detail for dataset preparation. Setup Runtime Environment Nvidia Jetson Orin + CUDA 11. 11 Compile && Run FP16 Performance && Metrics Average perf in FP16 on the training set(7481 instances) of KITTI dataset. 3D moderate metrics on the validation set(3769 instances
) of KITTI dataset. Note Voxelization has random output since GPU processes all points simultaneously while points selection for a voxel is random. References Detecting Objects in Point Clouds with NVIDIA CUDA-Pointpillars PointPillars: Fast Encoders for Object Detection from Point Clouds
Frequently asked questions
How many stars and forks are recorded?
NVIDIA-AI-IOT/CUDA-PointPillars has 659 stars and 185 forks.
What is the repository rank?
NVIDIA-AI-IOT/CUDA-PointPillars is ranked 15,366.
What language and category are recorded?
NVIDIA-AI-IOT/CUDA-PointPillars has 659 stars, uses Python and belongs to Model Development.
Which repository is closest in rank?
NVIDIA-AI-IOT/CUDA-PointPillars is compared with MorvanZhou/pytorch-A3C, ranked 15,365 with 659 stars.
PointPillars Inference with TensorRT
This repository contains sources and model for pointpillars inference using TensorRT.
Overall inference has below phases:
- Voxelize points cloud into 10-channel features
- Run TensorRT engine to get detection feature
- Parse detection feature and apply NMS
Prerequisites
Prepare Model && Data
We provide a Dockerfile to ease environment setup. Please execute the following command to build the docker image after nvidia-docker installation:
cd docker && docker build . -t pointpillar
We can then run the docker with the following command:
nvidia-docker run --rm -ti -v /home/$USER/:/home/$USER/ --net=host --rm pointpillar:latest
For model exporting, please run the following command to clone pcdet repo and install custom CUDA extensions:
git clone https://github.com/open-mmlab/OpenPCDet.git
cd OpenPCDet && git checkout 846cf3e && python3 setup.py develop
Download PTM to ckpts/, then use below command to export ONNX model:
python3 tool/export_onnx.py --ckpt ckpts/pointpillar_7728.pth --out_dir model
Use below command to evaluate on kitti dataset, follow Evaluation on Kitti to get more detail for dataset preparation.
sh tool/evaluate_kitti_val.sh
Setup Runtime Environment
- Nvidia Jetson Orin + CUDA 11.4 + cuDNN 8.9.0 + TensorRT 8.6.11
Compile && Run
sudo apt-get install git-lfs && git lfs install
git clone https://github.com/NVIDIA-AI-IOT/CUDA-PointPillars.git
cd CUDA-PointPillars && . tool/environment.sh
mkdir build && cd build
cmake .. && make -j$(nproc)
cd ../ && sh tool/build_trt_engine.sh
cd build && ./pointpillar ../data/ ../data/ --timer
FP16 Performance && Metrics
Average perf in FP16 on the training set(7481 instances) of KITTI dataset.
| Function(unit:ms) | Orin |
| ----------------- | ------ |
| Voxelization | 0.18 |
| Backbone & Head | 4.87 |
| Decoder & NMS | 1.79 |
| Overall | 6.84 |
3D moderate metrics on the validation set(3769 instances) of KITTI dataset.
| | Car@R11 | Pedestrian@R11 | Cyclist@R11 |
| ----------------- | --------| -------------- | ------------ |
| CUDA-PointPillars | 77.00 | 52.50 | 62.26 |
| OpenPCDet | 77.28 | 52.29 | 62.68 |
Note
- Voxelization has random output since GPU processes all points simultaneously while points selection for a voxel is random.
References