• 📬 Primary contact: Tianyu Li ( [email protected] )
• arXiv paper | OpenReview, ICLR 2024
• Blog (Zhihu) | Slides TODO

🔥 We advocate Lane Segment as a map learning paradigm that seamlessly incorporates both map 🛣️ geometry and 🕸️ topology information.

🏁 Lane Segment and OpenLane-V2 Map Element Bucket serve as a track in the CVPR 2024 Autonomous Grand Challenge.

• Official website: 🌐 AGC2024 Mapless Driving
• Evaluation server: 🤗 Hugging Face

This repository can be used as a starting point for Mapless Driving track.

• [2024/3] We released a multi-head version of LaneSegNet for the OpenLane-V2 Map Element Bucket!
  • Refer to this config file to train the baseline!
  • Please use the *_ls.pkl files generated by OpenLane-V2 preprocess scripts!
  • You can use ./tools/dist_test.sh {GPUs} --format-only to generate the submission.pkl now!
• [2023/12] LaneSegNet paper is available on arXiv. Code is also released!

• Model Zoo
• Prerequisites
• Installation
• Prepare Dataset
• Train and Evaluate
• License and Citation

The mean AP is between lane segment and pedestrian crossing.

This is a naive multi-branch model for the Map Element Bucket.
The pedestrian and road boundary are detected by an additional MapTR head. The traffic element are detected by a Deformable DETR head. The hyper-parameters are roughly set.

• Linux
• Python 3.8.x
• NVIDIA GPU + CUDA 11.1
• PyTorch 1.9.1

We recommend using conda to run the code.

conda create -n lanesegnet python=3.8 -y
conda activate lanesegnet

# (optional) If you have CUDA installed on your computer, skip this step.
conda install cudatoolkit=11.1.1 -c conda-forge

pip install torch==1.9.1+cu111 torchvision==0.10.1+cu111 -f https://download.pytorch.org/whl/torch_stable.html

Install mm-series packages.

pip install mmcv-full==1.5.2 -f https://download.openmmlab.com/mmcv/dist/cu111/torch1.9.0/index.html
pip install mmdet==2.26.0
pip install mmsegmentation==0.29.1
pip install mmdet3d==1.0.0rc6

Install other required packages.

pip install -r requirements.txt

Following OpenLane-V2 repo to download the Image and the Map Element Bucket data. Run the following script to collect data for this repo.

cd LaneSegNet
mkdir data

ln -s {Path to OpenLane-V2 repo}/data/OpenLane-V2 ./data/
python ./tools/data_process.py

After setup, the hierarchy of folder data is described below:

data/OpenLane-V2
├── train
|   └── ...
├── val
|   └── ...
├── test
|   └── ...
├── data_dict_subset_A_train_lanesegnet.pkl
├── data_dict_subset_A_val_lanesegnet.pkl
├── ...

We recommend using 8 GPUs for training. If a different number of GPUs is utilized, you can enhance performance by configuring the --autoscale-lr option. The training logs will be saved to work_dirs/lanesegnet.

cd LaneSegNet
mkdir -p work_dirs/lanesegnet

./tools/dist_train.sh 8 [--autoscale-lr]

You can set --show to visualize the results.

./tools/dist_test.sh 8 [--show]

All assets and code are under the Apache 2.0 license unless specified otherwise.

If this work is helpful for your research, please consider citing the following BibTeX entry.

@inproceedings{li2023lanesegnet,
  title={LaneSegNet: Map Learning with Lane Segment Perception for Autonomous Driving},
  author={Li, Tianyu and Jia, Peijin and Wang, Bangjun and Chen, Li and Jiang, Kun and Yan, Junchi and Li, Hongyang},
  booktitle={ICLR},
  year={2024}
}

@inproceedings{wang2023openlanev2,
  title={OpenLane-V2: A Topology Reasoning Benchmark for Unified 3D HD Mapping}, 
  author={Wang, Huijie and Li, Tianyu and Li, Yang and Chen, Li and Sima, Chonghao and Liu, Zhenbo and Wang, Bangjun and Jia, Peijin and Wang, Yuting and Jiang, Shengyin and Wen, Feng and Xu, Hang and Luo, Ping and Yan, Junchi and Zhang, Wei and Li, Hongyang},
  booktitle={NeurIPS},
  year={2023}
}

We acknowledge all the open-source contributors for the following projects to make this work possible:

• Openlane-V2
• BEVFormer
• TopoNet
• PersFormer
• OpenLane
• MapTR