PiFeNet: Accurate and Real-time 3D Pedestrian Detection Using an Efficient Attentive Pillar Network

Official implementation for KITTI/JRDB object detection

Citation

If you find this repo useful, please consider citing us, appreciate it!

@article{le2022accurate,
  title={Accurate and Real-time 3D Pedestrian Detection Using an Efficient Attentive Pillar Network},
  author={Le, Duy Tho and Shi, Hengcan and Rezatofighi, Hamid and Cai, Jianfei},
  journal={IEEE Robotics and Automation Letters},
  year={2022},
  publisher={IEEE}
}

Introduction

This repository is based on SECOND project.

ONLY support python 3.6+, pytorch 1.0.0+. Tested in Ubuntu 16.04/18.04/20.04/22.04, Windows 10.

This repo is not optimal on nuScenes dataset, consider using Det3D

If you want to train nuScenes dataset, see this.

News

12/2022: The paper has been accepted for publication at IEEE Robotics and Automation Letters (RA-L)

02/2022: JRDB dataset supported

WARNING: you should rerun info generation after every code update.

Performance in KITTI test set

PiFeNet/KITTI/xyres_16_submission.config + 150 epochs:

Benchmark	                Easy   Moderate	 Hard
Pedestrian (Detection)	        72.74%	62.35%	59.29%
Pedestrian (Orientation)	55.11%	46.59%	44.14%
Pedestrian (3D Detection)	56.39%	46.71%	42.71%
Pedestrian (Bird's Eye View)	63.25%	53.92%	50.53%

Performance in JRDB test set

PiFeNet/jrdb22/xyres_16_largea_JRDB2022.config + 40 epochs:

JRDB 2019:

         [email protected]      [email protected]     [email protected]
PiFeNet  74.284      42.617      4.886

JRDB 2022:

         [email protected]      [email protected]     [email protected]
PiFeNet  70.724      39.838      4.59

Install

1. Clone code

git clone https://github.com/ldtho/PiFeNet.git --recursive
cd ./PiFeNet/second

2. Install dependence python packages

It is recommend to use Anaconda package manager.

Create environment:

conda create --name pifenet python=3.8.6 pytorch=1.7.1 cudatoolkit=11.0.221 cudatoolkit-dev cmake=3.18.2 cuda-nvcc cudnn boost -c pytorch -c conda-forge -c nvidia
conda activate pifenet

Install dependencies

conda install addict einops fire jupyterlab jupyter-packaging tensorboard libboost matplotlib numba numpy open3d addict scikit-image psutil boost einops scikit-learn fire jupyterlab tensorboardx libboost matplotlib numba numpy open3d pandas pillow protobuf scipy seaborn tqdm yaml -c pytorch -c conda-forge -c nvidia -c numba -c open3d-admin
pip install opencv-python 

Follow instructions in spconv to install spconv or you can try:

git clone https://github.com/traveller59/spconv.git --recursive
cd spconv
git checkout fad3000249d27ca918f2655ff73c41f39b0f3127 && git submodule update --recursive    
python setup.py bdist_wheel
cd dist && pip install .

Sample conda environment PiFeNet.yml is available for your reference.

If you want to train with fp16 mixed precision (train faster in RTX series, Titan V/RTX and Tesla V100, but I only have 1080Ti), you need to install apex.

4. add second.pytorch/ to PYTHONPATH

export PYTHONPATH=$PYTHONPATH:/path/to/PiFeNet

Prepare dataset

• KITTI Dataset preparation

Download KITTI dataset and create some directories first:

└── KITTI_DATASET_ROOT
       ├── training    <-- 7481 train data
       |   ├── image_2 <-- for visualization
       |   ├── calib
       |   ├── label_2
       |   ├── velodyne
       |   └── velodyne_reduced <-- empty directory
       └── testing     <-- 7580 test data
           ├── image_2 <-- for visualization
           ├── calib
           ├── velodyne
           └── velodyne_reduced <-- empty directory

Then run

python create_data.py kitti_data_prep --root_path=KITTI_DATASET_ROOT

• JRDB Dataset preparation.

Download NuScenes dataset:

└── train_dataset_with_activity
       ├── calibration       
       ├── images          <-- frames without annotation
       ├── detections      <-- sample test detection - unused
       └── pointclouds     <-- point cloud files
└── test_dataset_without_labels
       ├── calibration       
       ├── images          <-- frames without annotation
       ├── detections      <-- sample train detection - unused
       ├── labels          <-- train set annotations
       └── pointclouds     <-- point cloud files

Then run

python create_data.py jrdb22_data_prep --root_path=JRDB_DATASET_ROOT

• Modify config file

There is some path need to be configured in config file:

train_input_reader: {
  ...
  database_sampler {
    database_info_path: "/path/to/dataset_dbinfos_train.pkl"
    ...
  }
  dataset: {
    dataset_class_name: "DATASET_NAME"
    kitti_info_path: "/path/to/dataset_infos_train.pkl"
    kitti_root_path: "DATASET_ROOT"
  }
}
...
eval_input_reader: {
  ...
  dataset: {
    dataset_class_name: "DATASET_NAME"
    kitti_info_path: "/path/to/dataset_infos_val.pkl"
    kitti_root_path: "DATASET_ROOT"
  }
}

Usage

Train with single GPU

KITTI:

python ./pytorch/train.py train --config_path=./configs/PiFeNet/KITTI/xyres_16_submission.config --model_dir=/your/save/dir

JRDB:

python ./pytorch/train.py train --config_path=./configs/PiFeNet/jrdb/xyres_16_largea_JRDB2022.config --model_dir=/your/save/dir

Note: add --resume=True if you want to continue training where you left off.

train with multiple GPU (need test, I only have one GPU)

Assume you have 4 GPUs and want to train with 3 GPUs:

CUDA_VISIBLE_DEVICES=0,1,3 python ./pytorch/train.py train --config_path=./configs/PiFeNet/KITTI/xyres_16_submission.config --model_dir=/your/save/dir --multi_gpu=True

Note: The batch_size and num_workers in config file is per-GPU, if you use multi-gpu, they will be multiplied by number of GPUs. Don't modify them manually.

You need to modify total step in config file. For example, 50 epochs = 15500 steps for car.lite.config and single GPU, if you use 4 GPUs, you need to divide steps and steps_per_eval by 4.

train with fp16 (mixed precision)

Modify config file, set enable_mixed_precision to true.

• Make sure "/path/to/model_dir" doesn't exist if you want to train new model. A new directory will be created if the model_dir doesn't exist, otherwise will read checkpoints in it.

• training process use batchsize=2 as default, you can increase or decrease depend on your GPU capability

evaluate

python ./pytorch/train.py evaluate --config_path=./configs/PiFeNet/KITTI/xyres_16_submission.config --model_dir=/your/save/dir --measure_time=True --batch_size=1

• detection result will saved as a result.pkl file in model_dir/eval_results/step_xxx or save as official KITTI label format if you use --pickle_result=False.

pretrained model

You can download pretrained models in google drive. The models' configs are the same as above

Try Kitti Viewer Web

Major step

1. run python ./kittiviewer/backend/main.py main --port=xxxx in your server/local.

2. run cd ./kittiviewer/frontend && python -m http.server to launch a local web server.

3. open your browser and enter your frontend url (e.g. http://127.0.0.1:8000, default]).

4. input backend url (e.g. http://127.0.0.1:16666)

5. input root path, info path and det path (optional)

6. click load, loadDet (optional), input image index in center bottom of screen and press Enter.

Inference step

Firstly the load button must be clicked and load successfully.

1. input checkpointPath and configPath.

2. click buildNet.

3. click inference.

Try Kitti Viewer (Deprecated)

You should use kitti viewer based on pyqt and pyqtgraph to check data before training.

run python ./kittiviewer/viewer.py, check following picture to use kitti viewer:

JRDB visualisation

First run the model, make predictions and convert is to KITTI-similar format

python ./pytorch/train.py evaluate --config_path=/path/to/config --model_dir=/pretrained_model/location --ckpt_path=/path/to/pretrained_model.tckpt

Follow the instructions in the JRDB visualisation toolkit

Result:

I also share the JRDB test set detections in KITTI format so that you can test the visualisation script.

2D:

3D:

Acknowledgement

• second.pytorch