ezKernel / MFNet

The code of MFNet: A Novel GNN-Based Multi-Level Feature Network With Superpixel Priors

MFNet

The code of MFNet: A Novel GNN-Based Multi-Level Feature Network With Superpixel Priors

Pipeline

Datasets

Directory structure reference SODData.py

• DUTS
• ECSSD
• PASCAL-S
• DUT-OMRON
• HKU-IS
• SOD
• THUR
• MSRA10K
• SED

Installing

torch==1.4.0+cu100

pip install torch-scatter==latest+cu100 -f https://pytorch-geometric.com/whl/torch-1.4.0.html
pip install torch-sparse==latest+cu100 -f https://pytorch-geometric.com/whl/torch-1.4.0.html
pip install torch-cluster==latest+cu100 -f https://pytorch-geometric.com/whl/torch-1.4.0.html
pip install torch-spline-conv==latest+cu100 -f https://pytorch-geometric.com/whl/torch-1.4.0.html
pip install torch-geometric=1.4.3

Training

1. Download VGG pretrained model: vgg16-397923af.pth

2. Set dataset path: _data_root_path = "/your/data/path/SOD/DUTS"

3. Just Run.

Eval

1. Set Param

model_name = "MFNet_SOD"
from MFNet_SOD import MyGCNNet, MyDataset
model_file = "./ckpt/MFNet_SOD/0/epoch_xx.pkl"
_data_root_path = "/your/data/path/SOD"  # Change to Your Data Path
result_path = "./result/MFNet_SOD/0/epoch_xx"

2. Just Run.

3. For Final Eval Result, Please Use: https://github.com/ArcherFMY/sal_eval_toolbox

Result

Reference

@ARTICLE{9953965,
  author={Li, Shuo and Liu, Fang and Jiao, Licheng and Chen, Puhua and Liu, Xu and Li, Lingling},
  journal={IEEE Transactions on Image Processing}, 
  title={MFNet: A Novel GNN-Based Multi-Level Feature Network With Superpixel Priors}, 
  year={2022},
  volume={31},
  number={},
  pages={7306-7321},
  doi={10.1109/TIP.2022.3220057}
}