zhangbaijin / memorynet Goto Github PK

Code for paper: Memory Augment is All Your Need for image restoration(cloud,rain,shadow removal, low-light image enhancement, image deblur)即插即用提点的记忆模块

License: MIT License

Python 86.20% MATLAB 13.80%

shadow-removal cloud-removal image-restoration pytorch image-deblurring rain-removal

memorynet's Introduction

MemoryNet:Memory augument is All You Need for image restoration

Memory augument is All You Need for image restoration Memory是一个即插即用的记忆模块，随时提点！ Paper link https://arxiv.org/abs/2309.01377

The structure of MemoryNet

Results of Shadow removal on ISTD dataset

Quick Run

To test the pre-trained models of Decloud on your own images, run

python demo.py --task Task_Name --input_dir path_to_images --result_dir save_images_here

Pretrained model

Download the pretrained model cloud-removal

2.Baidu Drive: 链接：https://pan.baidu.com/s/1nBNEsRLIFS2VVtHl8O14Rw 提取码：5mli

Dataset

Download datasets RICE from here, and ISTD dataset from here

To reproduce PSNR/SSIM scores of the paper, run MATLAB script

evaluate_PSNR_SSIM.m

ACKNOLAGEMENT

The code is updated on https://github.com/swz30/MPRNet

memorynet's People

Contributors

Stargazers

Watchers

Forkers

shumile66 nanjingxiaobawang mxsurui zlx1134558955 huangwenwenlili dv-lwd ip-restoration kelisidan1

memorynet's Issues

关于训练的问题

您好，运行环境和原作者的运行环境是一样的吗？
在训练的时候是否出现了下面这种问题：

In MemoryNet.py, MemoryNet() forward(). How to get the x2 and x3, thif function input is x3_img

运行后报错，找不到文件
PS C:\Users\dell\Desktop\yhdd\test_cloud\MemoryNet-main> python demo.py --task Deraining --input_dir ./Test-image-RICE/cloudy_image --result_dir ./result/1.png
Traceback (most recent call last):
File "C:\Users\dell\Desktop\yhdd\test_cloud\MemoryNet-main\demo.py", line 52, in
load_file = run_path(os.path.join(task, "MPRNet.py"))
File "D:\yhdd\language\python\lib\runpy.py", line 288, in run_path
code, fname = _get_code_from_file(run_name, path_name)
File "D:\yhdd\language\python\lib\runpy.py", line 252, in _get_code_from_file
with io.open_code(decoded_path) as f:
FileNotFoundError: [Errno 2] No such file or directory: 'C:\Users\dell\Desktop\yhdd\test_cloud\MemoryNet-main\Deraining\MPRNet.py'

您好，预训练去云模型网盘地址失效，可否提供新的地址，谢谢！

demo的model是？

你好！我想直接用您train好的model来试试，但遇到了一些问题，还请指教！

To test the pre-trained models of Decloud on your own images, run

python demo.py --task Task_Name --input_dir path_to_images --result_dir save_images_here

这里必须指定的task，是Deblurring, Denoising和Deraining任选一个吗？它们对应的model是您提供的哪一个呢？我download下来的有三个model_best.pth, model_epoch_250_250.pth, model_latest.pth，不知该如何和这些task对应。谢谢！

FileNotFoundError: [Errno 2] No such file or directory: 'C:\\Users\\Waqas\\Downloads\\Constractive-learning-shadow-removal-main\\Constractive-learning-shadow-removal-main\\Deblurring\\ MPRNet.py'

FileNotFoundError: [Errno 2] No such file or directory: 'C:\Users\Waqas\Downloads\Constractive-learning-shadow-removal-main\Constractive-learning-shadow-removal-main\Deblurring\
MPRNet.py'

请问对应的论文是什么

好像没搜到本代码对应的论文

Increase model capacity

Hi @zhangbaijin,

Thank you for your fantastic works!
I have tested your model on a customized dataset and it works very well. However, I realized two things:

Doing the inference on resized image yielded much better results compared to inferencing on high resolution image using sliceding window. Though, during the training, you already used random cropping. Do you have any idea on this?
It seems that this archecture does not work well with low-lightning images (please correct me if I'm wrong).

Additionally, I want to know if there is anyway to increase the model capacity? As well as changing the size of the receptive field?
Any guidance is priceless.
Thank you so much!