Grid-Wise Drivable Area (DA) Detection using LiDAR

Grid-Wise DA Repository (provided by AVELab) is the open LiDAR Drivable Area (DA) detection frameworks that provides a open-source big-scale dataset with wide range of driving scenarios in an urban environment. This tools also provide Drivable Area Detection using LiDAR called Grid-DATrNet first grid-wise DA deection using LiDAR leveraging attention mechanism through Transformer.

See our explanation and analysis of grid-wise DA detection on publication : See the Unseen : Grid-Wise Drivable Area Detection using LiDAR Dataset and Network in Remote Sensing Journal 2024.

Experiment Requirements

We experiment grid-wise DA detection using system as below :

• Operating System Linux 20.4
• Python 3.8
• Torch 2.0
• CUDA 11.7
• GPU NVIDA RTX3700

Preparing Argoverse-grid Dataset

First We proposed open-source big-scale grid-wise DA detection using challenging Dataset Argoverse dataset leveraging HD Map provided in Argoverse dataset. Hence first download the official Argoverse 1 dataset and HD Map information from official Argoverse 1 Dataset download.

Then order the Argoverse 1 Dataset with file order as below.

tracking_train1_v1.1
      ├── argoverse-tracking
            ├── train1
                  ├── 0ef28d5c-ae34-370b-99e7-6709e1c4b929
                        ├── [All LiDAR sensor information of scene  0ef28d5c-ae34-370b-99e7-6709e1c4b929]
                  ├── 2bc6a872-9979-3493-82eb-fb55407473c9
                        ├── [ All LiDAR Sensor information of scene 2bc6a872-9979-3493-82eb-fb55407473c9]
                  
                  .....
                  
                  

There will be total 65 scenes folder for Argoverse 1 Dataset training

Then please set up the Argoverse 1 HD Map folder as below.

hdmaps
      ├── map_files
            ├── MIA_10316_driveable_area_mat_2019_05_28.npy
            ├── MIA_10316_ground_height_mat_2019_05_28.npy
            ...

Visualization of converting Argoverse 1 Dataset into Argoverse-grid can be seen as below.

Experiment Installation

Then please install modified Argoverse 1 API in the repository and install needed package for experiment Grid-Wise DA Detection as below.

1. First install modified Argoverse 1 API using command as below.

cd argoverse-api/
pip install -e .

2. Then install needed package for experiment Grid-Wise DA detection using LiDAR with running command as below.

cd ..
pip install -r requirements.txt

Experiment using Grid-DATrNet

We proposed novel grid-wise DA detection using attention mechanism through Transformer Grid-DATrNet. Architecture of grid-wise DA detection using attention mechanism can be seen as below.

Then We can experiment training using proposed Grid-DATrNet by set up the configs file and run the command as below.

1.Set up the dataset training path in config file. Change value dataset_path and dataset_path_val in line 130 and 131 in config file /configs/Grid-DATrNet_using_Global_Attention.py based on the Argoverse 1 training dataset path.

2.Then running the training python script using command as below.

sudo python train_gpu_0.py

Making DA Detection using Grid-DATrNet

For predicting DA Detection using Grid-DATrNet you can change the path of Grid-DATrNet model in variable path_ckpt in file validate_gpu_0.py based on your Grid-DATrNet model checkpoints and run testing python script using command as below.

sudo python validate_gpu_0.py

If you want to visualize DA detection result you can change value of setting is_visualized_result and is_save_visualization to True in configs file of Grid-DATrNet

Visualization of result proposed Grid-DATrNet on proposed Argoverse-Grid dataset can be seen as below.

Model Zoo

Name	Accuracy	F1 Score	GFLOPs	Model	Paper
Grid-DATrNet (PointPillar + Global Attention)	93.40	0.8321	180	Link	Link
Grid-DATrNet (PointPillar + MLP Mixer)	91.40	0.8145	110	Link	Link

License

Grid-Wise DA is released under the Apache-2.0 license.

Acknowledgement

The Grid-Wise DA benchmark is contributed by Christofel Rio Goenawan and Dong-Hee Paek, advised by Seung-Hyun Kong.

We thank the maintainers of the following projects that enable us to develop Grid-Wise DA: OpenPCDet by MMLAB, TuRoad bu TuZheng. K-Lane by Dong-Hee.

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. 2021R1A2C3008370).

Citation

If you find this work is useful for your research, please consider citing:

@article{goenawan2024see,
  title={See the Unseen: Grid-Wise Drivable Area Detection Dataset and Network Using LiDAR},
  author={Goenawan, Christofel Rio and Paek, Dong-Hee and Kong, Seung-Hyun},
  journal={Remote Sensing},
  volume={16},
  number={20},
  pages={3777},
  year={2024},
  doi={10.3390/rs16203777}
}