LDNet: End-to-End Lane Marking Detection Approach Using a Dynamic Vision Sensor

This is the Pytorch implementation of LDNet a Lane marking detection algorithm on DVS data. Papar "LDNet: End-to-End Lane Marking Detection Approach Using a Dynamic Vision Sensor", accepted in 2021.

Installation

To run the demo example you need only Pytorch, Numpy, and dropblock dependecies.

Main Dependencies:

• Pytorch 1.6.0
• Torchvision 0.7.0
• pyyaml 3.13
• DropBlock

Inorder to use this code you must install Anaconda and then apply the following steps:

• Create the environment from the environment.yml file:

conda env create -f environment.yml

• Activate LDNet environment

source activate LDNet

• Install DropBlock

pip install dropblock

cd LDNet/

pip install -r requirements.txt

Training

Dataset

Before start training, download DET dataset from here

Training paramteres

training.yaml contains parameters needed for training as:

• DATASET, path to dataset folder. The folder (DET) must follow this pattern

  /path/to/DET
              /train		
              /val 	  
              /test 
  

  Start training:

In order to train the network with a specific backbone network and get and replicate the paper result you must train network.

• To train on DET data set and run the training via

python train.py

Citation

@ARTICLE{9518365,
  author={Munir, Farzeen and Azam, Shoaib and Jeon, Moongu and Lee, Byung-Geun and Pedrycz, Witold},
  journal={IEEE Transactions on Intelligent Transportation Systems}, 
  title={LDNet: End-to-End Lane Marking Detection Approach Using a Dynamic Vision Sensor}, 
  year={2021},
  volume={},
  number={},
  pages={1-17},
  doi={10.1109/TITS.2021.3102479}}