OSCNET+ 代码复现-CSDN博客

本文链接：https://blog.csdn.net/buganything/article/details/140271808

项目github 已有，开个博客大家如果复现有问题可以随时在下面留言

github ：GitHub - hongwang01/OSCNet: 【MICCAI 2022, TMI 2023】Orientation-Shared Convolution Representation Model

1、从github 下载项目并解压

2、下载数据集

当然自己用肯定是自己写dataload，我就不放自己的代码了

3、然后训练就行

Train OSCNet

python train.py --gpu_id 0 --data_path "data/train/"  --model "osc" --log_dir "logs_osc/" --model_dir "pretrained_model/model_osc/" --cdiv 1

Train OSCNet+

python train.py --gpu_id 0 --data_path "data/train/"  --model "oscplus" --log_dir "logs_oscplus/" --model_dir "pretrained_model/model_oscplus/" --cdiv 10

4、想直接测试也有提供权重

python test.py  --gpu_id 0 --data_path "data/test/" --model "osc" --model_dir "pretrained_model/model_osc/net_latest.pt" --save_path "save_results/" --cdiv 1

python test.py  --gpu_id 0 --data_path "data/test/" --model "oscplus" --model_dir "pretrained_model/model_oscplus/net_latest.pt" --save_path "save_results/" --cdiv 10

Reconstruction results can be downloaded from [NetDisk](pwd:osc1)

@inproceedings{wang2022orientation,
title={Orientation-Shared Convolution Representation for CT Metal Artifact Learning},
author={Wang, Hong and Xie, Qi and Li, Yuexiang and Huang, Yawen and Meng, Deyu and Zheng, Yefeng},
booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
pages={665--675},
year={2022}