rolandying / FusionLane

The source code of paper FusionLane

FusionLane:

This is the source code of paper FusionLane: Multi-Sensor Fusion for Lane Marking Semantic Segmentation Using Deep Neural Networks. The source code of this project is modified based on deeplabV3+.

Setup

Requirements:

• Tensorflow>= 1.12
• Numpy
• matplotlib
• opencv-python

Dataset

This project uses the TFRecord format to consume data in the training and evaluation process. The TFRecord used in this project could is available at here. You can also find the testing images at here.

Training and Evaluation

You can start your own training from scratch as follows. Or you could just modified in the train.py file.

python  train.py -- data_dir TFRECORD_DATA_DIR \
                 -- model_dir MODEL_DIR

You can evaluate you model in the same way as the training. Our model is available at here.

---

Citation

If you feel our work is helpfull, please cite as follow:

@article{yin2020fusionlane,
  title={Fusionlane: Multi-sensor fusion for lane marking semantic segmentation using deep neural networks},
  author={Yin, Ruochen and Cheng, Yong and Wu, Huapeng and Song, Yuntao and Yu, Biao and Niu, Runxin},
  journal={IEEE Transactions on Intelligent Transportation Systems},
  volume={23},
  number={2},
  pages={1543--1553},
  year={2020},
  publisher={IEEE}
}