Parallel Heterogeneous Networks with Adaptive Routing for Online Video Lane Detection

Requirements

• PyTorch >= 1.10
• CUDA >= 10.0
• CuDNN >= 7.6.5
• python >= 3.6

Installation

1. Download repository. We call this directory as ROOT:

$ git clone https://github.com/CZY-Code/PHNet.git

2. Install dependencies:

$ conda install pytorch==1.10.0 torchvision==0.11.0 torchaudio==0.10.0 cudatoolkit=10.2 -c pytorch
$ pip install -r requirements.txt

3. Install NMS ops

$ cd ./libs/ops
$ python setup.py build develop

4. Install evaluation tools:

$ sudo ln -s /usr/local/include/opencv4/opencv2 /usr/local/include/opencv2$ 
$ cd evaluation/culane
$ mkdir build && cd build
$ cmake ..
$ make
$ mv culane_evaluator ../

Pytorch can be installed on here. Other versions might be available as well.

Dataset

1. Download OpenLane-V and VIL-100 dataset.
2. Unzip and move dataset into ROOT/dataset

Directory structure of dataset

├── OpenLane                
│   ├── images              
│   ├── OpenLane-V
|   |   ├── label          
|   |   ├── list            
├── VIL-100
│   ├── JPEGImages          
│   ├── Annotations
|   └── ...

Train and Test

0. Modify the settings in option files in ./options/option*.py
1. For OpenLane-V dataset:

$ torchrun --nproc_per_node=4 trainOL.py
$ python testOLV3.py

2. For VIL-100 dataset:

$ torchrun --nproc_per_node=4 trainVIL.py
$ python testVIL.py

Weights

Weight for OpenLane-V: [here](通过网盘分享的文件：50.pth.tar 链接: https://pan.baidu.com/s/15RTib-EKtDbYn7VFi13MHA?pwd=b4ac 提取码: b4ac).

Weight for VIL-100: [here](通过网盘分享的文件：50.pth.tar 链接: https://pan.baidu.com/s/1LBEEQMBKZb8sBeSZYOUTdg?pwd=jkba 提取码: jkba).

Evaluation

$ cd ./evaluation
$ python evaluate_iou.py #(VIL-100)
$ python evaluate_iou4OL.py #(OpenLane-V)

Acknowledgement

• dongkwonjin/RVLD
• yujun0-0/MMA-Net
• lucastabelini/LaneATT
• Turoad/CLRNet