Hi, thanks for the effort in the great challenge and dataset. I have a question about the implementation of compute_tooth_size function.
In the evaluation.py, the tooth size is computed as:
def compute_tooth_size(points, centroid):
size = np.sqrt(np.sum((centroid - points) ** 2, axis=0))
return size
but it seems different from the concept of actual physical size in mm, and the size is affected by the count of GT tooth points as it sums up all the points from the GT tooth.
I'm wondering is it designed as so, or the size could be calculated in another way like:
def proposed_compute_tooth_size(points, centroid):
size = np.sqrt(np.max((centroid - points) ** 2, axis=0)) * 2 # double the length from the centroid to the farthest boundingbox point
return size
To validate this, I choose a GT tooth from a sample model in the Teeth3DS dataset, and compute its size. The result from the current function is [390.28815167 432.65123712 226.6848727 ], while the result from the proposed function is [12.74475451 12.48835577 10.1212647 ].
Thanks for your reply in advance.
Hi, thanks for the effort in the great challenge and dataset. I have a question about the implementation of
compute_tooth_sizefunction.In the
evaluation.py, the tooth size is computed as:but it seems different from the concept of actual physical size in
mm, and the size is affected by the count of GT tooth points as it sums up all the points from the GT tooth.I'm wondering is it designed as so, or the size could be calculated in another way like:
To validate this, I choose a GT tooth from a sample model in the Teeth3DS dataset, and compute its size. The result from the current function is
[390.28815167 432.65123712 226.6848727 ], while the result from the proposed function is[12.74475451 12.48835577 10.1212647 ].Thanks for your reply in advance.