First, install python by conda,
conda create -n segmentation_DRAC python=3.9 -y
conda activate segmentation_DRAC
Then, install environment by pip,
pip install -r requirements.txt
Mix the training dataset and the testing dataset download from DRAC Challenge into one package like the following format.
data
├── A. Segmentation
│ ├── 1. Original Images
│ │ ├── a. Training Set
│ │ └── b. Testing Set
│ └── 2. Groundtruths
└── spilt_val.py
If data is already, please run spilt_val.py like below command.
cd data
python spilt_val.py
Then, data preparation is over.
Before training or testing the model, please open config.py to change the config which you use during training and testing.
Just run train.py or train.sh, like below command,
python train.py
or
bash train.sh
Run test.py
vis.py is to product prediction in order to view or submit.
ensemble.py is to ensemble model and product prediction in order to submit