gcn-parking-slot's Introduction

Attentional Graph Neural Network for Parking Slot Detection

Repository for the paper "Attentional Graph Neural Network for Parking Slot Detection".

@article{gcn-parking-slot:2020,
  title={Attentional Graph Neural Network for Parking Slot Detection},
  author={M. Chen, J. Xu, L. Xiao, D. Zhao etal},
  journal={IEEE Robotics and Automation Letters (RA-L)},
  year={2021},
  volume={6},
  number={2},
  pages={3445-3450},
  doi={10.1109/LRA.2021.3064270}
}

Requirements

python 3.6
pytorch 1.4+
other requirements: pip install -r requirements.txt

Pretrained models

Two pre-trained models can be downloaded with following links.

Link	Code	Description
Model0	bc0a	Trained with ps2.0 subset as in [1]
Model1	pgig	Trained with full ps2.0 dataset

Prepare data

The original ps2.0 data and label can be found here. Extract and organize as follows:

├── datasets
│   └── parking_slot
│       ├── annotations
│       ├── ps_json_label 
│       ├── testing
│       └── training

Train & Test

Export current directory to PYTHONPATH:

export PYTHONPATH=`pwd`

demo

python3 tools/demo.py -c config/ps_gat.yaml -m cache/ps_gat/100/models/checkpoint_epoch_200.pth

train

python3 tools/train.py -c config/ps_gat.yaml

test

python3 tools/test.py -c config/ps_gat.yaml -m cache/ps_gat/100/models/checkpoint_epoch_200.pth

References

[1] J. Huang, L. Zhang, Y. Shen, H. Zhang, and Y. Yang, “DMPR-PS: A novel approach for parking-slot detection using directional marking-point regression,” in IEEE International Conference on Multimedia and Expo (ICME), 2019. code

Recommend Projects