Images and annotations do not match
baifanxxx opened this issue · comments
Hi,
Thank you for your contribution. I loaded the ULS23_DeepLesion3D subset from your link. I found 750 images but 743 masks. In ULS23_Radboudumc_Bone, I also found 744 images but 697 masks. This means that the images and annotations do not match. Is it right? Or do I have something wrong?
Best regards,
BAI Fan
Hi BAI Fan,
You are correct that there are fewer masks than images in those cases. We removed/improved a number of incorrect training cases in this commit. You can discard the images that do not have a mask associated with them. At a later date, we will reorganize the image data on Zenodo and remove these cases.
Thank you for getting back to me. Your dataset is very helpful.