TopoNet: A New Baseline for Scene Topology Reasoning

Graph-based Topology Reasoning for Driving Scenes

• Production from OpenDriveLab at Shanghai AI Lab. Jointly with collaborators at Huawei.
• Primary contact: Tianyu Li ( [email protected] )

---

This repository contains the source code of TopoNet, Graph-based Topology Reasoning for Driving Scenes.

TopoNet is the first end-to-end framework capable of abstracting traffic knowledge beyond conventional perception tasks, i.e., reasoning connections between centerlines and traffic elements from sensor inputs. It unifies heterogeneous feature learning and enhances feature interactions via the graph neural network architecture and the knowledge graph design.

Instead of recognizing lanes, we adhere that modeling the lane topology is appropriate to construct road components within the perception framework, to facilitate the ultimate driving comfort. This is in accordance with the UniAD philosophy.

Table of Contents

• News
• Main Results
• Prerequisites
• Installation
• Prepare Dataset
• Train and Evaluate
• License
• Citation

News

• Pinned: The leaderboard for Lane Topology Challenge is open for regular submissions year around. This Challenge would be back in 2024's edition.
• [2023/11] 🔥The code and model of OpenLane-V2 subset-B is released!
• [2023/08] The code and model of TopoNet is released!
• [2023/04] TopoNet paper is available on arXiv.
• [2023/01] Introducing Autonomous Driving Challenge for Lane Topology at CVPR 2023.

Main Results

Results on OpenLane-V2 subset-A val

We provide results on Openlane-V2 subset-A val set.

Method	Backbone	Epoch	DETl	TOPll	DETt	TOPlt	OLS
STSU	ResNet-50	24	12.7	0.5	43.0	15.1	25.4
VectorMapNet	ResNet-50	24	11.1	0.4	41.7	6.2	20.8
MapTR	ResNet-50	24	8.3	0.2	43.5	5.8	20.0
MapTR*	ResNet-50	24	17.7	1.1	43.5	10.4	26.0
TopoNet	ResNet-50	24	28.6	4.1	48.6	20.3	35.6

🔥: Based on the updated v1.1 OpenLane-V2 devkit and metrics, we have reassessed the performance of TopoNet and other SOTA models. For more details please see issue #76 of OpenLane-V2.

Method	Backbone	Epoch	DETl	TOPll	DETt	TOPlt	OLS
STSU	ResNet-50	24	12.7	2.9	43.0	19.8	29.3
VectorMapNet	ResNet-50	24	11.1	2.7	41.7	9.2	24.9
MapTR	ResNet-50	24	8.3	2.3	43.5	8.9	24.2
MapTR*	ResNet-50	24	17.7	5.9	43.5	15.1	31.0
TopoNet	ResNet-50	24	28.6	10.9	48.6	23.8	39.8

*: evaluation based on matching results on Chamfer distance.
The result of TopoNet is from this repo.

Results on OpenLane-V2 subset-B val

Method	Backbone	Epoch	DETl	TOPll	DETt	TOPlt	OLS
TopoNet	ResNet-50	24	24.4	6.7	52.6	16.7	36.0

The result is based on the updated v1.1 OpenLane-V2 devkit and metrics.
The result of TopoNet is from this repo.

Model Zoo

Model	Dataset	Backbone	Epoch	OLS	Memory	Config	Download
TopoNet-R50	subset-A	ResNet-50	24	39.8	12.3G	config	ckpt / log
TopoNet-R50	subset-B	ResNet-50	24	36.0	8.2G	config	ckpt / log

Prerequisites

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

Installation

We recommend using conda to run the code.

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

# (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

Prepare Dataset

Following OpenLane-V2 repo to download the data and run the preprocessing code.

cd TopoNet
mkdir data && cd data

ln -s {PATH to OpenLane-V2 repo}/data/OpenLane-V2

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

data/OpenLane-V2
├── train
|   └── ...
├── val
|   └── ...
├── test
|   └── ...
├── data_dict_subset_A_train.pkl
├── data_dict_subset_A_val.pkl
├── data_dict_subset_B_train.pkl
├── data_dict_subset_B_val.pkl
├── ...

Train and Evaluate

Train

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/toponet.

cd TopoNet
mkdir -p work_dirs/toponet

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

Evaluate

You can set --show to visualize the results.

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

License

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

Citation

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

@article{li2023toponet,
  title={Graph-based Topology Reasoning for Driving Scenes},
  author={Li, Tianyu and Chen, Li and Wang, Huijie and Li, Yang and Yang, Jiazhi and Geng, Xiangwei and Jiang, Shengyin and Wang, Yuting and Xu, Hang and Xu, Chunjing and Yan, Junchi and Luo, Ping and Li, Hongyang},
  journal={arXiv preprint arXiv:2304.05277},
  year={2023}
}

@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}
}

Related resources

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

• Openlane-V2
• BEVFormer