weichen582/retinexnet

A Tensorflow implementation of RetinexNet

retinexnet's Issues

can not find decomposition results

i tried to set --decom=0 as well as --decom=1, but in both setup, i can not find the decomposed results, only the enhanced ones.

I_delta is the enhanced I_low ?

I_delta is the enhanced I_low ? think u

Run python main.py --phase=test.The image in the test_result can not open.BMP image has bogus header data.How to fix it?

Supplementary material is not available

Thanks for your job!
As the title says, however, i cannot download the supplmentary material in your project page anymore.
Could you update it or realease a new link?

Problem with dataset

Run the code and find that there is a problem with loading the dataset path

pretrained model

is there any pretrained model on large dataset we can use ?

error in LOLdataset

there are several mistakes on the correspondance in "our485", like the 79.png, 735.png.

Poor performance on my own data set

Denoising operation code

hi, can you include your denoising operation code?

License

Hi. I used the library and it works great! Could you please add a license to the repository?

Inference results seems weird using default trained model

I tried to follow the steps to generate enhanced images, but the results seems weird. Is that normal?

Inference results: https://drive.google.com/drive/folders/1TlEPm40HJS-A7qwD5ikY6DfEleG19d0T?usp=sharing

so much noise in the test_results compared the original one

The enhanced image has a lot of noise, although it has a little brighter than the original image.

ValueError: Variable RelightNet/conv2d/kernel already exists, disallowed. in"I_delta = RelightNet(I_low, R_low)"

when i run the main.py for train, i got "ValueError: Variable RelightNet/conv2d/kernel already exists, disallowed. Did you mean to set reuse=True or reuse=tf.AUTO_REUSE in VarScope? Originally defined at:" in line68 "I_delta = RelightNet(I_low, R_low)" of model.py

somebody can helps me???

version of python

It shows up some errors when I ran the code. The error is "Process finished with exit code -1073741819 (0xC0000005)". I don't think the version of the python I used is right. Could you please tell me the version of yours? Thanks.

数据集光照强度有具体数值嘛

请问有数据集光照强度的具体值嘛

why RelightNet output not sigmoid?

训练Decomposition模型时出现的问题

1. 训练时LOSS值一直稳定不变，应该和learning_rate没什么关系，无论怎么调整都不会变。
2. Decomposition模型输出的R_low和I_low会随着训练次数的增加逐渐变白，最后变成全白色。
   这两个是什么原因？我是单独训练Decomposition，应该和这个没关系吧。

合成的数据集下载后为什么是brightening train的数据集？能分享下真正的合成数据集吗？谢谢

Could I ask which code for psnr, ssim calculation?

Is it possible to release the evaluation code for fair comparison? Thank you.

About the synthetic dataset

I noticed that there are some deformations in the synthetic image pairs (384×384) provided by your project page. Maybe you did some operations like resize? Do they have any impacts on your results?
By the way, in your paper, you said that these raw images are resized to 400×600. Why are they 384×384 currently?

Too many compile error

my pytorch version is 1.1.0

1. in dataset.py
   Line 51: class TheDataset(data.dataset):
   class TheDataset(data.dataset): TypeError: module() takes at most 2 arguments (3 given)
   I solve it by changing to:class TheDataset(data.Dataset)

2.in train.py
Line 64: dataloader = torch.utils.data.dataLoader(train_set, batch_size=args.batch_size...)
dataloader = torch.utils.data.dataLoader(train_set, batch_size=args.batch_size, shuffle=True, AttributeError: module 'torch.utils.data' has no attribute 'dataLoader'
I solve it by changing to:dataloader = torch.utils.data.DataLoader(train_set, batch_size=args.batch_size...)

diffMAE

Dear author, hello. I recently read your paper "diffmae" and found it very interesting. Can you make your code publicly available?

Nice for sharing, howerver, Poor performance even using images in data/eval

感谢分享，
但是，
CPU模式下，对data/eval /test下结果也不好，噪点非常明显

python3.6 main.py --use_gpu=0 --phase=test --test_dir=/home//下载/RetinexNet-master/test1/ --save_dir=/home//下载/RetinexNet-master/res1/ --decom=1
是还有其他参数吗？
而且decom参数貌似没用啊

Use of reduce_max in input

In the beginning of DeconNet function, there are these two ines:

input_max = tf.reduce_max(input_im, axis=3, keepdims=True)
input_im = concat([input_max, input_im])

Could you tell what it does and what is the use of the same?

run "python man.py --phase=test" and produced no results