Lanenet-Lane-Detection (pytorch version)

Introduction

Use pytorch to implement a Deep Neural Network for real time lane detection mainly based on the IEEE IV conference paper "Towards End-to-End Lane Detection: an Instance Segmentation Approach".You can refer to their paper for details https://arxiv.org/abs/1802.05591. This model consists of ENet encoder, ENet decoder for binary semantic segmentation and ENet decoder for instance semantic segmentation using discriminative loss function.

The main network architecture is:

Configure the environment

conda create -n pytorch1.7.0 python=3.7
conda activate pytorch1.7.0
conda install pytorch==1.7.0 torchvision==0.8.0 torchaudio==0.7.0 cudatoolkit=11.0 -c pytorch
pip install -r requirements.txt

Generate Tusimple training set/validation set

First, download tusimple dataset here through the kagge website.

path/to/your/unzipped/file should like this:

|--train_set
|----clips
|----label_data_0313.json
|----label_data_0531.json
|----label_data_0601.json
|--test_set
|----...
|--test_label.json

Then, run the following command to generate the training folder and the train.txt,val.txt file.

python tusimple_transform.py --src_dir ./TUSimple/train_set --val True

Training the model

Using tusimple folder with ENet/Focal loss:

python train.py --dataset ./TUSimple/train_set/training

You cloud also replace ENet and Cross Entropy with DeepLabv3+ or Cross Entropy loss through following commands

python train.py --dataset ./TUSimple/train_set/training --loss_type CrossEntropyLoss

python train.py --dataset ./TUSimple/train_set/training --model_type DeepLabv3+

If you want to change focal loss to cross entropy loss, do not forget to adjust the hyper-parameter of instance loss and binary loss in ./model/lanenet/train_lanenet.py

Testing result

A pretrained trained model by myself is located in ./log (only trained in 25 epochs)
Test the model:

python test.py --img ./data/tusimple_test_image/0.jpg --model_type ENet --model ./log/best_model.pth

The testing result is here:

Evalution

python eval.py --dataset ./TUSimple/train_set/training --model_type ENet --model ./log/best_model.pth

Reference:

The project is modified from https://github.com/IrohXu/lanenet-lane-detection-pytorch.