JK-the-Ko / SAR-CAM

[IEEE JSTARS] Official PyTorch Implementation of "SAR Image Despeckling Using Continuous Attention Module"
MIT License
29 stars 4 forks source link

Dataset issues #4

Open sun668899 opened 1 year ago

sun668899 commented 1 year ago

Could you please send me the adjusted data set? I would be very grateful.

JK-the-Ko commented 1 year ago

Hello, the GPU server that I have used is currently broken. Hence, I will not be able to hand you processed dataset. Thank you.

sun668899 commented 1 year ago

Is there any code for processing the dataset?

JK-the-Ko commented 1 year ago

I do not have the code. Please use Python and OpenCV to extract patches and use Matlab for adding speckle noise.

Polaris119 commented 1 year ago

Hello, after I use Matlab to add speckle noise, the training result loss is a negative value. What is the reason?

JK-the-Ko commented 1 year ago

As stated in the paper, we have proposed a modified DG loss which has a log scale. Hence, if the DG becomes less than one, the modified DG loss returns a negative value. Please refer to the paper for more details.

Akhte10 commented 8 months ago

C:\ProgramData\Anaconda3\python.exe D:\code\1\SAR-CAM-main\SAR-CAM-main\test.py usage: test.py [-h] [--model-name MODEL_NAME] --weights-dir WEIGHTS_DIR --clean-image-dir CLEAN_IMAGE_DIR --noisy-image-dir NOISY_IMAGE_DIR --save-dir SAVE_DIR [--stack-image] [--device DEVICE] test.py: error: the following arguments are required: --weights-dir, --clean-image-dir, --noisy-image-dir, --save-dir

Process finished with exit code 2

JK-the-Ko commented 8 months ago

Hi, you need to specifically add path of the corresponding arguments to run the code.