search

Bobholamovic / CDLab

Yet another repository for developing and benchmarking deep learning-based change detection methods.
The Unlicense
188 stars 30 forks source link

The label of the CDD　dataset is not binary map. #5

Closed yiyi-today closed 2 years ago

yiyi-today commented 2 years ago

Thank you for your generous sharing！

I downloaded the CDD dataset mentioned in your paper, but found that the label of the CDD dataset was not binary map, as Issue in label dataset of 2018. Lebedev M A, Vizilter Y V, Vygolov O V, et al. Change detection in remote sensing images using conditional adversarial networks said.

How did you deal with this situation ?

thanks

Bobholamovic commented 2 years ago

I binarize the labels with 127 as the threshold. See https://github.com/Bobholamovic/CDLab/blob/4cab42ebe60a95203e5b67f6680947e9a559eed6/src/data/svcd.py#L50

yiyi-today commented 2 years ago

I binarize the labels with 127 as the threshold. See

https://github.com/Bobholamovic/CDLab/blob/4cab42ebe60a95203e5b67f6680947e9a559eed6/src/data/svcd.py#L50

I binarize the labels with 127 as the threshold. See

https://github.com/Bobholamovic/CDLab/blob/4cab42ebe60a95203e5b67f6680947e9a559eed6/src/data/svcd.py#L50

Thanks a lot for your answer, but what is the special meaning of using 127? If you normalize all values to 0-1, there will be many noisy points, is this normal? 1653618462(1)

Bobholamovic commented 2 years ago

I binarize the labels with 127 as the threshold. See https://github.com/Bobholamovic/CDLab/blob/4cab42ebe60a95203e5b67f6680947e9a559eed6/src/data/svcd.py#L50

I binarize the labels with 127 as the threshold. See https://github.com/Bobholamovic/CDLab/blob/4cab42ebe60a95203e5b67f6680947e9a559eed6/src/data/svcd.py#L50

Thanks a lot for your answer, but what is the special meaning of using 127? If you normalize all values to 0-1, there will be many noisy points, is this normal? 1653618462(1)

  1. The numeric range of uint8 type is [0,255], while 127 is (0+255)//2.
  2. A binarization result like this might be normal, depending on the threshold you use.
yiyi-today commented 2 years ago

I binarize the labels with 127 as the threshold. See https://github.com/Bobholamovic/CDLab/blob/4cab42ebe60a95203e5b67f6680947e9a559eed6/src/data/svcd.py#L50

I binarize the labels with 127 as the threshold. See https://github.com/Bobholamovic/CDLab/blob/4cab42ebe60a95203e5b67f6680947e9a559eed6/src/data/svcd.py#L50

Thanks a lot for your answer, but what is the special meaning of using 127? If you normalize all values to 0-1, there will be many noisy points, is this normal? 1653618462(1)

  1. The numeric range of uint8 type is [0,255], while 127 is (0+255)//2.
  2. A binarization result like this might be normal, depending on the threshold you use.

Thank you and have a nice life！