About Datasets

[M0ck4ry]

Hello, I noticed that in your code, the training and testing sets are placed in the same folder. So, during testing, a portion of the images are randomly selected for segmentation testing? Or will the last 100 images be used for testing? If you could reply, I would greatly appreciate it

[bravo-hq]

Dear @M0ck4ry,

Thank you for contacting us. In the datasets directory, we have outlined how the data is split for each dataset.

For example, in the case of the PH2 dataset, the code specifies the split as follows:

• The first 80 images are used for training.
• The next 20 images are used for validation.
• The subsequent 100 images are used for testing.

This means the dataset split is not done randomly; it follows a predefined order as mentioned above.

If you have any further questions or need additional clarification, please feel free to ask.

Sincerely, Yousef Sadegheih

[M0ck4ry]

Dear @bravo-hq , Thank you for your reply. Yes, I noticed what you said. But I used 10 sample images for testing, the final segmentation corresponding to the original image is different from the 10 sample images I found. What is the reason for this? If I could receive your reply, I would greatly appreciate it

[bravo-hq]

Dear @M0ck4ry,

Could you please provide more details about the dataset you are working on and the specific changes you have made? This information will help us assist you more effectively. It's important to ensure that the pairs of images and labels remain together and are not mismatched.

Sincerely, Yousef Sadegheih

[M0ck4ry]

Dear @bravo-hq Thank you for your reply again. I tried to train with 80 images, validate with 10 images, and test with 9 images (these 99 images are located in a separate folder). It was found that the 9 images tested were completely different from the last 9 images in the previous dataset.Here are some screenshots and images of the files The first image is the log section, and the second image is the generated result. I found that these 9 test images are completely different from the last 9 of the 99 images I set (i.e. the test set). Normally, im.png and gt.png should be identical to the previous pictures. But the result is not like that. I would greatly appreciate it if you could reply

[Burger-Z]

Does ISIC2016 dataset work on your code? Because ISIC2018 is too large and I cannot successfully download it.