Skip to main content Skip to navigation

Q&A

glas3resize2.png

1. Q: Why do I not receive a confirmation email after registration?

A: After registration, please allow at least 24 hours before we can send you an email with the log-in and password details required for downloading the data. If you do not receive any response from us after 3 days, please register again.


2. Q: We have downloaded the training data set, but looking at the annotated images they look to me not so clear. Sometimes they are completely black, some other times with very smooth grey shades.

A: These are all label images. Each pixel in an individual glandular object is labelled with a unique number, and the background pixel is labelled with zero. If there are 10 objects in an image then unique intensity values are 0,1,…,10. The reason that they look dim is that they were saved as an 8-bit grey scale image (intensity range = 0-255).


3. Q: There are some images where a large portion of the background is white because of the background slide. I wanted to know whether we could fill that region with any other color or should we proceed with the original images and not change them at all?

A: Background removal should be incorporated as part of your segmentation pipeline, and should not be performed manually.


4. Q: Is the final score the average across all testing images?

A: The final score is not an average. It will be calculated from all test images by considering them as part of a bigger image. However, in the short paper, you can report an average score on the training dataset.

jwyns3zjnew.png