当前位置:网站首页>代码学习(DeamNet)CVPR | Adaptive Consistency Prior based Deep Network for Image Denoising
代码学习(DeamNet)CVPR | Adaptive Consistency Prior based Deep Network for Image Denoising
2022-07-17 06:21:00 【Claire_Shang】
本文提出了一种新的深度网络图像去噪方法。 与现有的基于深度网络的去噪方法不同,我们将新的ACP项引入到优化问题中,然后利用优化过程,通过展开策略设计深度网络。 我们的ACP驱动去噪网络结合了经典去噪方法的一些有价值的成果,并在一定程度上提高了其可解释性。
官方代码:

1 介绍
### Contents
Dataset|包含三个文件夹(train, test和Benchmark_test),您可以将train数据集放在train中,将测试数据集放在test中,将SIDD/DnD基准放在Benchmark_test中
Deam_models| 当您想测试|时,预先训练的模型
real| 一些关于真实图像去噪的python文件
statistics| 训练时记录结果
### Training on AWGN
要重新训练我们的网络,请将您自己的训练数据集放在'./Dataset/train ',然后运行' train.py’
### Training on real-world noise
重新训练我们的网络以去除真实世界的噪声: 将训练数据集下载到`./Dataset/train`并用'./Dataset/train/gen_dataset_real.py'将它们打包成h5py格式。
你可以从https://www.eecs.yorku.ca/~kamel/sidd/dataset.php
和http://ani.stat.fsu.edu/~abarbu/Renoir.html获取数据集
将训练和测试路径设置为您自己的路径,并运行' train.py '。更多详细信息,请参考https://github.com/JimmyChame/SADNet
### Testing on AWGN
要测试你自己的图像,把你的数据集放在' Dataset/test/your_test_name '中,然后运行' Synthetic_test.py '
### Testing on real-world noise
为了测试真实世界的噪声数据集:将测试数据集下载到'./Dataset/Benchmark_test '并运行
'Benchmark_test.py '
您可以从 https://www.eecs.yorku.ca/~kamel/sidd/benchmark.php and https://noise.visinf.tu-darmstadt.de/benchmark/ 获得数据集.
----------------------------------------------------------------------------------
先运行train.py
D:\ProgramData\Anaconda3\envs\python36\python.exe "D:/Papers to read/2022.07/Adaptive Consistency Prior based Deep Network for Image_Denoising/DeamNet-main/DeamNet-main/train.py"
Traceback (most recent call last):
File "D:/Papers to read/2022.07/Adaptive Consistency Prior based Deep Network for Image_Denoising/DeamNet-main/DeamNet-main/train.py", line 15, in <module>
from real_dataloader import *
File "D:\Papers to read\2022.07\Adaptive Consistency Prior based Deep Network for Image_Denoising\DeamNet-main\DeamNet-main\real_dataloader.py", line 6, in <module>
import h5py
ModuleNotFoundError: No module named 'h5py'
Process finished with exit code 1
报错原因:安装h5py库
下载数据库,选了一个最小的(都要1.6GB). 如果运行成功,下一步准备学习‘Dataset.py’,结合代码学习一下文章内容。
边栏推荐
- Pytorch随记(2)
- High performance integrated video image processing board based on ultrascale FPGA + Huawei Hisilicon arm / fpga+arm
- Is there any cumulative error in serial communication and the requirements for clock accuracy
- Redis jump table implementation principle & time complexity analysis
- @How can conditionalonmissingbean cover beans in third-party components
- Pycharm interface settings
- How does Jenkins set the mailbox to automatically send mail?
- The best storage scheme of MCU CS Chuangshi SD NAND
- Using PCA to simplify data
- 导出文件or下载文件
猜你喜欢
随机推荐
CCF-CSP《202206-2—寻宝!大冒险!》
912. 排序数组(数组排序)
Machine learning interview questions (Reprinted)
MongoDB的下载、安装和使用
Redis storage structure principle 2
INSTALL_PARSE_FAILED_MANIFEST_MALFORMED
Virtual machine stack of [JVM]
How to choose the right model
牛客题目——打家劫舍一、打家劫舍二
Pytorch随记(1)
A set of Jenkins style for personal use
并发编程的核心问题
How does Jenkins set the mailbox to automatically send mail?
SCI论文的Highlights怎么写(正经的教你怎么写)
Pytoch notes (1)
导出文件or下载文件
《牛客刷题》sql错题集
正则表达示提取匹配内容
【MySQL】 事务:事务基础知识、MySQL事务实现原理、事务日志 redolog & undolog 详解
Ku115 FPGA high performance 10G Optical fiber network hardware accelerator card / 2-way 10G Optical fiber data accelerator card









