Introduction

This repository is an official implementation of SQD-MapNet.

Getting Started

1. Environment

Step 1. Create conda environment and activate it.

conda create --name sqdmapnet python=3.8 -y
conda activate sqdmapnet

Step 2. Install PyTorch.

pip install torch==1.9.0+cu111 torchvision==0.10.0+cu111 torchaudio==0.9.0 -f https://download.pytorch.org/whl/torch_stable.html

Step 3. Install MMCV series.

# Install mmcv-series
pip install mmcv-full==1.6.0
pip install mmdet==2.28.2
pip install mmsegmentation==0.30.0
git clone https://github.com/open-mmlab/mmdetection3d.git
cd mmdetection3d
git checkout v1.0.0rc6 
pip install -e .
pip install -r requirements.txt

Step 4. Install other requirements.

cd ..
pip install -r requirements.txt

2. Data Preparation

Step 1. Download NuScenes dataset to ./datasets/nuScenes.

Step 2. Download Argoverse2 (sensor) dataset to ./datasets/av2.

Step 3. Generate annotation files for NuScenes dataset.

python tools/nuscenes_converter.py --data-root ./datasets/nuScenes

Step 4. Generate annotation files for Argoverse2 dataset.

python tools/argoverse_converter.py --data-root ./datasets/av2

3. Training and Validating

To train a model with 8 GPUs:

bash tools/dist_train.sh ${CONFIG} 8

To validate a model with 8 GPUs:

bash tools/dist_test.sh ${CONFIG} ${CEHCKPOINT} 8 --eval

Results

Results on NuScenes

Range	$\mathrm{AP}_{ped}$	$\mathrm{AP}_{div}$	$\mathrm{AP}_{bound}$	$\mathrm{AP}$	Config	Epoch	Checkpoint
$60\times 30\ m$	63.0	65.5	63.3	63.9	Config	24	ckpt
$100\times 50\ m$	67.0	65.5	59.5	64.0	Config	24	ckpt

📖 Citation

If you find our code or paper helps, please consider citing:

@article{wang2024stream,
  title={Stream query denoising for vectorized hd map construction},
  author={Wang, Shuo and Jia, Fan and Liu, Yingfei and Zhao, Yucheng and Chen, Zehui and Wang, Tiancai and Zhang, Chi and Zhang, Xiangyu and Zhao, Feng},
  journal={arXiv preprint arXiv:2401.09112},
  year={2024}
}

Acknowledgments

We sincerely thank the open-sourcing of these works where our code is based on: StreamMapNet and DN-DETR.