by Yi Wang, Zijun Deng, Xiaowei Hu, Lei Zhu, Xin Yang, Xuemiao Xu, Pheng-Ann Heng, and Dong Ni

This implementation is written by Zijun Deng at the South China University of Technology.

@inproceedings{wang18d,
     author = {Wang, Yi and Deng, Zijun and Hu, Xiaowei and Zhu, Lei and Yang, Xin and Xu, Xuemiao and Heng, Pheng-Ann and Ni, Dong},
     title = {Deep Attentional Features for Prostate Segmentation in Ultrasound},
     booktitle = {MICCAI},
     year = {2018}
}

• Python 2.7
• PyTorch >= 0.3.0
• torchvision
• numpy
• Cython
• pydensecrf (here to install)
• training set and testing set (you need to prepare them by yourself and groundtruths should be binary masks)

1. Set the path of pretrained ResNeXt model in resnext/config.py
2. Set the path of training set in config.py
3. Run by python train.py

The pretrained ResNeXt model is ported from the official torch version, using the convertor provided by clcarwin. You can directly download the pretrained model ported by me.

Hyper-parameters of training were gathered at the beginning of train.py and you can conveniently change them as you need.

Training a model on a single GTX 1080Ti GPU takes about 20 minutes.

1. Set the path of testing set in config.py
2. Run by python infer.py