different results with paper's

[mmmmli]

Is there something different with the paper?

[mmmmli]

Hello, I just run this code with your settings, but the results are quite different with the paper's result.

[Angus1996]

I trained the network, but the result is not good. The PSNR and SSIM of pred result are both lower the the LDCT. Did you meet the same problem?

[IronLavender]

Hello, is there anyone who can send me the dataset to me please? I send a request following the email address on the website, but have no reply. And I'm kind of in a hurry to do the experiments. Thank you very much. My email is: ml.wu@siat.ac.cn

[xiaorannuo]

I trained the network, but the result is not good. The PSNR and SSIM of pred result are both lower the the LDCT. Did you meet the same problem?

Did you find out which code was wrong?I have the same question.

[fxyjj]

I trained the network, but the result is not good. The PSNR and SSIM of pred result are both lower the the LDCT. Did you meet the same problem?

Did you find out which code was wrong?I have the same question.

I also got the same problem with the wrong result. When i was testing my model, the predict images are totally black, while the measurements (PSNR & SSIM) indicates that the result should not be black, quite confusion. I guess there may be some problem at the image showing steps or somewhere else...

[zxzbuaa]

I can't run the code. It shows 'ValueError: num_samples should be a positive integer value, but got num_samples=0'. What about you?

[zxzbuaa]

I trained the network, but the result is not good. The PSNR and SSIM of pred result are both lower the the LDCT. Did you meet the same problem?

Which kind of data or pictures did you use?

[fxyjj]

I can't run the code. It shows 'ValueError: num_samples should be a positive integer value, but got num_samples=0'. What about you?

you can try to figure out whether there are images in the test file (in this code, the test file is set to be L506, I got the same question as you and I thought it is because of the empty test file.

[fxyjj]

Hello, is there anyone who can send me the dataset to me please? I send a request following the email address on the website, but have no reply. And I'm kind of in a hurry to do the experiments. Thank you very much. My email is: [ml.wu@siat.ac.cn](mailt

I trained the network, but the result is not good. The PSNR and SSIM of pred result are both lower the the LDCT. Did you meet the same problem?

hi, I tried this code too, and I also got the same situation, while in my opinion, there maybe some hardware limitations causing the difference, and the results is not the same when run the code firstly and secondly, and I also think it is the unstable training process of the discriminator, as you can see, in GAN structure, Discriminator is difficult to train and there is why there is a gradient panelty (gp) parameter to optimize the training process of the Discriminator, even though gp exists, sometimes I also got the black result (possibly gradient vanishing or exploding happens during the training process of the Discriminator). welcome to share your ideas with me, fortunately I run this code successfully and got some similar result as the thesis showed, but the result is not what I want, so I am so willing to communicate with people working on this code!

[zxzbuaa]

I can't run the code. It shows 'ValueError: num_samples should be a positive integer value, but got num_samples=0'. What about you?

you can try to figure out whether there are images in the test file (in this code, the test file is set to be L506, I got the same question as you and I thought it is because of the empty test file.

I creat some zip files named like L506, and in which file put two folder named 'quarter_3mm' and 'full_3mm' in it, like the code said. Each folder have several pictures (and I also tried the dicom file). How did you build your data file?

[fxyjj]

Hello, is there anyone who can send me the dataset to me please? I send a request following the email address on the website, but have no reply. And I'm kind of in a hurry to do the experiments. Thank you very much. My email is: [ml.wu@siat.ac.cn](mailt
I trained the network, but the result is not good. The PSNR and SSIM of pred result are both lower the the LDCT. Did you meet the same problem?
hi, I tried this code too, and I also got the same situation, while in my opinion, there maybe some hardware limitations causing the difference, and the results is not the same when run the code firstly and secondly, and I also think it is the unstable training process of the discriminator, as you can see, in GAN structure, Discriminator is difficult to train and there is why there is a gradient panelty (gp) parameter to optimize the training process of the Discriminator, even though gp exists, sometimes I also got the black result (possibly gradient vanishing or exploding happens during the training process of the Discriminator). welcome to share your ideas with me, fortunately I run this code successfully and got some similar result as the thesis showed, but the result is not what I want, so I am so willing to communicate with people working on this code!

Thank you very much! Can you send an email to me so I can get your address? My address is: zxz1609@buaa.edu.cn

hi, I cannot share the dataset with you. You need to got to the website and contact the owner of the dataset, more details could bee found here: https://www.aapm.org/GrandChallenge/LowDoseCT/
it may cause sometime to get the dataset, if you are in a hurry, please do it asap, good luck!

[zxzbuaa]

Hello, is there anyone who can send me the dataset to me please? I send a request following the email address on the website, but have no reply. And I'm kind of in a hurry to do the experiments. Thank you very much. My email is: [ml.wu@siat.ac.cn](mailt
I trained the network, but the result is not good. The PSNR and SSIM of pred result are both lower the the LDCT. Did you meet the same problem?
hi, I tried this code too, and I also got the same situation, while in my opinion, there maybe some hardware limitations causing the difference, and the results is not the same when run the code firstly and secondly, and I also think it is the unstable training process of the discriminator, as you can see, in GAN structure, Discriminator is difficult to train and there is why there is a gradient panelty (gp) parameter to optimize the training process of the Discriminator, even though gp exists, sometimes I also got the black result (possibly gradient vanishing or exploding happens during the training process of the Discriminator). welcome to share your ideas with me, fortunately I run this code successfully and got some similar result as the thesis showed, but the result is not what I want, so I am so willing to communicate with people working on this code!

Thank you very much! Can you send an email to me so I can get your address? My address is: zxz1609@buaa.edu.cn

hi, I cannot share the dataset with you. You need to got to the website and contact the owner of the dataset, more details could bee found here: https://www.aapm.org/GrandChallenge/LowDoseCT/
it may cause sometime to get the dataset, if you are in a hurry, please do it asap, good luck!

I send email to them about two month ago but they told me they have many applications to handle...The time has been delayed for many times. Or, can you help me to run the code with my own pictures?

[fxyjj]

I can't run the code. It shows 'ValueError: num_samples should be a positive integer value, but got num_samples=0'. What about you?
you can try to figure out whether there are images in the test file (in this code, the test file is set to be L506, I got the same question as you and I thought it is because of the empty test file.

I creat some zip files named like L506, and in which file put two folder named 'quarter_3mm' and 'full_3mm' in it, like the code said. Each folder have several pictures (and I also tried the dicom file). How did you build your data file?

just like the readme file said, yes, when you uncompress the zip files, there are two files (full_3mm and quarter_3mm), with particular patients no.(L067, L096 ..), I mean the question you described is caused by the empty test file, (L506 is set to be the test file, and you can modified it if you like, in the main.py, there is an argument test_file = L506, you can try to modified it to L067, L096 or the file you have, honestly, I am not sure whether you have the similar situation as I had, since when I got such error, I realized that I make the test file empty by mistake. I suggest you to use just one series of images(e.g, just L067 ) to do your first running, convenience and easy to debug, time saving as well, and modified the training set and test set to be the same, and see whether there still has such error, hoping this is helpful!

[fxyjj]

Hello, is there anyone who can send me the dataset to me please? I send a request following the email address on the website, but have no reply. And I'm kind of in a hurry to do the experiments. Thank you very much. My email is: [ml.wu@siat.ac.cn](mailt
I trained the network, but the result is not good. The PSNR and SSIM of pred result are both lower the the LDCT. Did you meet the same problem?
hi, I tried this code too, and I also got the same situation, while in my opinion, there maybe some hardware limitations causing the difference, and the results is not the same when run the code firstly and secondly, and I also think it is the unstable training process of the discriminator, as you can see, in GAN structure, Discriminator is difficult to train and there is why there is a gradient panelty (gp) parameter to optimize the training process of the Discriminator, even though gp exists, sometimes I also got the black result (possibly gradient vanishing or exploding happens during the training process of the Discriminator). welcome to share your ideas with me, fortunately I run this code successfully and got some similar result as the thesis showed, but the result is not what I want, so I am so willing to communicate with people working on this code!
Thank you very much! Can you send an email to me so I can get your address? My address is: zxz1609@buaa.edu.cn
hi, I cannot share the dataset with you. You need to got to the website and contact the owner of the dataset, more details could bee found here: https://www.aapm.org/GrandChallenge/LowDoseCT/
it may cause sometime to get the dataset, if you are in a hurry, please do it asap, good luck!

I send email to them about two month ago but they told me they have many applications to handle...The time has been delayed for many times. Or, can you help me to run the code with my own pictures?

well, as you can see, it is not a simple work, I can tell you how the network works, but the image preprocessing process is difficult, moreover, I got the official dataset and run it successfully, so I just look through the preprocessing parts, I've spent much time on learning the network structure, so , I'm afraid I can help you very little...

[fxyjj]

Hello, is there anyone who can send me the dataset to me please? I send a request following the email address on the website, but have no reply. And I'm kind of in a hurry to do the experiments. Thank you very much. My email is: [ml.wu@siat.ac.cn](mailt
I trained the network, but the result is not good. The PSNR and SSIM of pred result are both lower the the LDCT. Did you meet the same problem?
hi, I tried this code too, and I also got the same situation, while in my opinion, there maybe some hardware limitations causing the difference, and the results is not the same when run the code firstly and secondly, and I also think it is the unstable training process of the discriminator, as you can see, in GAN structure, Discriminator is difficult to train and there is why there is a gradient panelty (gp) parameter to optimize the training process of the Discriminator, even though gp exists, sometimes I also got the black result (possibly gradient vanishing or exploding happens during the training process of the Discriminator). welcome to share your ideas with me, fortunately I run this code successfully and got some similar result as the thesis showed, but the result is not what I want, so I am so willing to communicate with people working on this code!
Thank you very much! Can you send an email to me so I can get your address? My address is: zxz1609@buaa.edu.cn
hi, I cannot share the dataset with you. You need to got to the website and contact the owner of the dataset, more details could bee found here: https://www.aapm.org/GrandChallenge/LowDoseCT/
it may cause sometime to get the dataset, if you are in a hurry, please do it asap, good luck!

I send email to them about two month ago but they told me they have many applications to handle...The time has been delayed for many times. Or, can you help me to run the code with my own pictures?

Hi, there. how is your work going? i did my dissertation based on this program, i remembered that you may need some help about process you own images right? maybe i can offer some help to you now! free to contact me and here is my email: fxyj123@yeah.net.
I'm so pleased to help you!

[young2157]

Hello, is there anyone who can send me the dataset to me please? I send a request following the email address on the website, but have no reply. And I'm kind of in a hurry to do the experiments. Thank you very much. My email is: [ml.wu@siat.ac.cn](mailt

I trained the network, but the result is not good. The PSNR and SSIM of pred result are both lower the the LDCT. Did you meet the same problem?

hi, I tried this code too, and I also got the same situation, while in my opinion, there maybe some hardware limitations causing the difference, and the results is not the same when run the code firstly and secondly, and I also think it is the unstable training process of the discriminator, as you can see, in GAN structure, Discriminator is difficult to train and there is why there is a gradient panelty (gp) parameter to optimize the training process of the Discriminator, even though gp exists, sometimes I also got the black result (possibly gradient vanishing or exploding happens during the training process of the Discriminator). welcome to share your ideas with me, fortunately I run this code successfully and got some similar result as the thesis showed, but the result is not what I want, so I am so willing to communicate with people working on this code!

Hello, may I ask how you solved the black image in the test? According to what you said, there was no gradient disappearance or explosion in the loss of discriminator in the network I observed, but the black image still appeared in the test. How did you solve it? Thank u so much

[asherp7]

Any progress with the black predictions issue?
The generator seems to constantly fall into predicting patches to be black, time after time.

[Taogonglin]

I trained the network, but the result is not good. The PSNR and SSIM of pred result are both lower the the LDCT. Did you meet the same problem?

I meet the same question, and when I test I get the black image ,and you? And I wonder know if you solve it. Could you share you solution with me?

[sbajh123]

Hello, I just run this code with your settings, but the results are quite different with the paper's result.

I also encountered the same problem, did you solve this problem? Thank you

[sbajh123]

I trained the network, but the result is not good. The PSNR and SSIM of pred result are both lower the the LDCT. Did you meet the same problem?

I also encountered the same problem, did you solve this problem? Thank you

[respect3]

你好，我只是用你的设置运行这段代码，但是结果与论文的结果有很大的不同。

我也遇到了同样的问题，请问你解决了这个问题吗？谢谢

Hi,could you please tell me your email address for better communication?And my email is wangxiangcanwww@163.com

[sbajh123]

你好，我只是用你的设置运行这段代码，但是结果与论文的结果有很大的不同。

我也遇到了同样的问题，请问你解决了这个问题吗？谢谢

Hi,could you please tell me your email address for better communication?And my email is wangxiangcanwww@163.com

my email is xhchun1@163.com

[respect3]

I trained the network, but the result is not good. The PSNR and SSIM of pred result are both lower the the LDCT. Did you meet the same problem?

I meet the same question, and when I test I get the black image ,and you? And I wonder know if you solve it. Could you share you solution with me?

Hi,could you please tell me your email address for better communication?And my email is wangxiangcanwww@163.com

[joaolcguerreiro]

The problem still persists... How can I replicate this paper?

[EvelyneCalista]

Hi, did anyone solve the black image in inference problem? i need a help.
thank you.