coordinates of bounding box of nodules
DSRajesh opened this issue · comments
Hello
I have a doubt regarding the discussion in
#132.
How can zmin, zmax take values 144, 199 respectively when there just 82 slices as shown in the question below
The ordering of the coordinates must be like zmin, ymin, zmax, ymax, xmin, xmax right ?
Also why the ordering is like this if it is so ?
Thank you
Rajesh
Dear Rajesh,
the format is always a bit tricky due to sitk..
In the following examples the axis ordering can be interchanged arbitrarily and all the predictions will follow the relativ orderings (that is why I prefer to use ax0, ax1, ax2 instead of the typical x,y,z .. )
Nifty Ordering of the Image/sitk image: z,y,x
Array ordering (after GetArrayFromImage): x,y,z
Predictions: x_min, y_min, x_max, y_max, z_min, z_max
In the above example (in the image):
Nifty ordering (not shown): x,y,z
Array ordering (this is seen in 'original_size_of_raw_data'): z,y,x
Predictions: z_min, y_min, z_max, y_max, x_min, x_max
-> note how the indices of the predcitions always follow the ordering of the array format; the first 4 entries reference the first two axis of the array and the last two entries refer to the last axes of the array
As indicated in my comment in the original post: given an image (I meant array there) in x,y,z format
The arrows in your post show from x,y,z predictions to an z,y,x array.
(hope this makes sense 😅 )
Regarding the format:
The x_min, y_min, x_max, y_max format is commonly used in object detection frameworks in the natural image computing domain. The reason to extend it in the way we did it, is code reuse from the natural image computing domain -> this allows for a simple extension of the third dimension without needing to rewrite entire functions.
This issue is stale because it has been open for 30 days with no activity.
This issue was closed because it has been inactive for 14 days since being marked as stale.