DLFNet: Multi-Scale Dynamic Weighted Lane Feature Network for Complex Scenes

Pytorch implementation of the paper "DLFNet: Multi-Scale Dynamic Weighted Lane Feature Network for Complex Scenes"(ICIC2025).

Introduction

• DLFNet is based on the BiFPN concept and the way humans perceive and reason about lane lines in the real world, achieving the integration of global semantic information with local feature details.
• In culane and tusimple, the performance is superior , especially at high IOU threshold

Installation

Prerequisites

Only test on Ubuntu18.04 and 20.04 with:

• Python >= 3.8 (tested with Python3.8)
• PyTorch >= 1.6 (tested with Pytorch1.8)
• CUDA (tested with cuda11.1)
• Other dependencies described in requirements.txt

Clone this repository

Clone this code to your workspace. We call this directory as $DLFNET_ROOT

git clone https://github.com/EADMO/DLFNet.git

Create a conda virtual environment and activate it (conda is optional)

conda create -n dlfnet python=3.8 -y
conda activate dlfnet

Install dependencies

# Install pytorch firstly, the cudatoolkit version should be same in your system.

conda install pytorch==1.8.0 torchvision==0.9.0 torchaudio==0.8.0 cudatoolkit=11.1 -c pytorch -c conda-forge

# Install python packages
python setup.py build develop

Data preparation

CULane

Download CULane. Then extract them to $CULANEROOT. Create link to data directory.

cd $DLFNET_ROOT
mkdir -p data
ln -s $CULANEROOT data/CULane

For CULane, you should have structure like this:

$CULANEROOT/driver_xx_xxframe    # data folders x6
$CULANEROOT/laneseg_label_w16    # lane segmentation labels
$CULANEROOT/list                 # data lists

Tusimple

Download Tusimple. Then extract them to $TUSIMPLEROOT. Create link to data directory.

cd $DLFNET_ROOT
mkdir -p data
ln -s $TUSIMPLEROOT data/tusimple

For Tusimple, you should have structure like this:

$TUSIMPLEROOT/clips # data folders
$TUSIMPLEROOT/lable_data_xxxx.json # label json file x4
$TUSIMPLEROOT/test_tasks_0627.json # test tasks json file
$TUSIMPLEROOT/test_label.json # test label json file

For Tusimple, the segmentation annotation is not provided, hence we need to generate segmentation from the json annotation.

python tools/generate_seg_tusimple.py --root $TUSIMPLEROOT
# this will generate seg_label directory

Getting Started

Training

For training, run

python main.py [configs/path_to_your_config] --gpus [gpu_num]

For example, run

python main.py configs/resnet18_culane.py --gpus 0

Validation

For testing, run

python main.py [configs/path_to_your_config] --[test|validate] --load_from [path_to_your_model] --gpus [gpu_num]

For example, run

python main.py configs/dla34_culane.py --validate --load_from culane_dla34.pth --gpus 0

At present, this code can output the visual results and GT . You only need to add add --view or --view_gt We will get the visualization result in work_dirs/xxx/xxx/visualization.

For generate DEMO, run

python generate_DEMO.py

Replace image_folder with your visual path

Results

CULane

Backbone	mF1	F1@50	F1@75
ResNet-18	55.35	79.77	62.64
ResNet-34	55.14	79.77	62.83
ResNet-101	55.86	80.18	63.43
DLA-34	56.25	80.45	63.56

TuSimple

Backbone	F1	Acc	FDR	FNR
ResNet-18	97.91	96.88	2.54	1.62
ResNet-34	97.82	96.92	2.56	1.79
ResNet-101	97.89	96.77	1.81	2.41

Acknowledgement

• Turoad/clrnet
• open-mmlab/mmdetection
• pytorch/vision
• Turoad/lanedet
• ZJULearning/resa
• cfzd/Ultra-Fast-Lane-Detection
• lucastabelini/LaneATT
• aliyun/conditional-lane-detection