You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hello
I have planned to use LNDb chest CT scan datasets for training a nodule detector model using your nice "nnDetection". Some of the images of these datasets do not have unit normal direction vectors. Also the dimensions of these images are such that there are more than 512 voxels along the x-axis. Simply meaning, dataset image sizes are something like 674x512x442, 632x512x416, etc where as usually the dataset images of "other than LNDb" datasets have typical size like 512x512x442, 512x512x416, etc, (i mean to say they will have same no of voxels along both x and y axes). Now my questions are:
Can I use them for training ?
If i can, do i need to apply some transforms to these images and then use them for training ?
It would be very helpful, if i could get answers for these questions. Thank you
Rajesh
The text was updated successfully, but these errors were encountered:
nnDetection automatically handles images without unit normal direction vectors (note: the vectors this need to be orthogonal otherwise SimpleITK will complain) and of varying images sizes, nothing special todo.
Hello
I have planned to use LNDb chest CT scan datasets for training a nodule detector model using your nice "nnDetection". Some of the images of these datasets do not have unit normal direction vectors. Also the dimensions of these images are such that there are more than 512 voxels along the x-axis. Simply meaning, dataset image sizes are something like 674x512x442, 632x512x416, etc where as usually the dataset images of "other than LNDb" datasets have typical size like 512x512x442, 512x512x416, etc, (i mean to say they will have same no of voxels along both x and y axes). Now my questions are:
It would be very helpful, if i could get answers for these questions. Thank you
Rajesh
The text was updated successfully, but these errors were encountered: