Car-Lane-Detection

This repo was designed to use in Baidu Car Lane Detection Challenge.

Preparation

1. Download the dataset from here
2. Setup the data folder in this structure

   Image/
     road02/
       Record002/
         Camera 5/
           ...
         Camera 6
       Record003
       ....
     road03
     road04
   Label/
     Label_road02/
      Label
       Record002/
         Camera 5/
          ...
         Camera 6
       Record003
       ....
     Label_road03
     Label_road04

3. Change the path in make_list.py and run

image_dir = Path('yourPath/Image_Data/')
label_dir = Path('yourPath/Label/')
save_path = Path('yourPath/data_list/')

python -make_list.py

After this step, train dataset, validate dataset will be created.

Train

1. Change the GPU setting in train.py

os.environ["CUDA_VISIBLE_DEVICES"] = '0'
device_list = [0]

2. Run train.py

python -train.py

Predict

After training, model will be stored under logs. Run predict.py

python -predict.py